Please subscribe my channel
Zebrainy ABC Wonderlands - Learn A to B alphabet letters - Education Game App for Kid
Zebrainy ABC Wonderlands - Learn A to B alphabet letters - Education Game App for Kid
Đăng ký:
Đăng Nhận xét (Atom)
-------------------------------------------
Understanding P.S.H.E (Personal Social and Health education) - Duration: 6:22.
So hello everybody this is me Asim and welcome to Asim and Aysha channel
Today we will be talking about PSHE, first of all we are gonna say what PSHE stands for it stands for Personal
Social, Health and Education
Personal about your own self social about the other
sociality that's around you, and you're
interacting with and health your health
Your daily needs like food shelter water
All your daily essentials so and all about that in one place called PSHE
education Woooo!!!!!
Okay
PSHE, just like
Trains us for the for when we get above 18 and 40 and stuff
So it just trains us it gives us support, so we so we don't
fall and
Collapse along then. I just hit my back, but don't don't care these things alright
okay, so
We're gonna talk about friends what's the importance of friends? How do we make friends never talk about friendship another strong?
These are the topics we're going to cover today, so let's get on whoa
so so
So I made this diagram here, and it's kind of notes, but I'm just gonna pace your hair
So y'all can understand
Happy bunch
So I was just like let's we had PSHE class on Tuesday, so I was like let's let's make
A chart some like topic PSEG, and I name the topic friends, so there are lots of topics like
respect
friends
other stuff
So here are some tips to have best friend
So I just want to explain how to be best friend. You know be lonely like a pathetic sad guy
So what is never spread rumors about your friend like there are lots of people that spread rumors about their friends
and that's wrong so don't spend rumors about and
be
Honest, I didn't ever like to mention a few friends
That would be sent from you and never there talk to you really fast right now, and why but I have so
If you're not your friend that just means that you're not a good you're not a good person
Just like them and showing off
Like you're like yeah, dude
I totally got that new Call of Duty game yeah, boyee oh
nice
Yeah boy, and when they come to you have noting
III don't have that sorry I lied, and that's when you're not a good friend so don't never do that stuff
But one is backbitting your friend is bad bad bad
It's a islamic term that backbitting is bad like you're like you don't bite other people's backs
It's not that thing
It's just like oh, it's also kind of spreading roomers
But in the end of the way, you're like yo done ready spirit is so
ugly oh really
It's kind of nice no one talks about that anymore this
2018
Fashion rule, I know what you say
Yeah, so about that stuff. I just made a little skit cut that out
So don't beat them ever
If you beat your friends like you're you're probably just cut just just just throw them out of the window
Find yourself other friends and if that
It's okay, so then I was like let's a little bit
Humidity or so
If you follow all these rules that congratulations you have a best friend whoa
How's bored swine draught some air sorry so you?
have to love
appreciate and be
Happy with your friends if you don't like love your friends are gonna be like no man
yeah, we've been friends for a really long time and
Sorry, but no no. I I just remembered something I was in school and
If you haven't watched my
100 bill subscribers video my super hundreds of couples mitral video chat are by the way and
So my friend she he got my coconut my fir chihiro's. It's wat so there's another
Friend of mine in my school, and he was like dude. I thought you mentioned mentioned you mean
He called mom and stuff, but where he like
Played the video I showed my friends picture
And then he was like dude when he saw when he said that she here
He's my best friend. I I thought that you're talking about me
He showed it to
Say that thank God not be Lolita forward, and when she took the picture like now you're so those
Fish that from it by the way. I'm saying that other
She here is my cousin, and we've been friends for real, and you're also my best friend like you support me
I'll be trying like I also give you a shout-out to it
It was check out red pepper white she that my friend shadow I first video got collaborate in collab with a move
So I'll get a link in a description and everything yeah
I don't know his logo his logo would be a pepper laughing at you
So we're going off topic so
Yeah, pshe
personal so sure
education
That's it
so if you liked this video sure
subscribe like my video like this video and share me everybody else so thank you guys for watching this video and
Like it subscribe for videos just like this one. Bye
-------------------------------------------
Zebrainy ABC Wonderlands - Learn M alphabet letters - Education Game App for Kid - Duration: 3:08.
Zebrainy ABC Wonderlands - Learn A to B alphabet letters - Education Game App for Kid
Zebrainy ABC Wonderlands - Learn A to B alphabet letters - Education Game App for Kid
-------------------------------------------
Marco Reus Lifestyle, Net Worth, Salary,House,Cars, Awards, Education, Biography And Family - Duration: 4:46.
-------------------------------------------
Life Hero Campaign for Thai Education - Duration: 2:06.
Guess in which category Thailand has made it to the Top 3 in the world?
Most tourists? Nope...
Fastest Development? Not that either...
Actually, Thailand is No. 3 in the world ranking of countries with the highest "Inequality."
The income gap between the rich and the poor keeps getting wider and wider...
because of unequal access to education.
Currently, as many as 430,000 youths in Thailand are out-of-school.
These kids tend to be from poor families and often drop out from lower secondary school,
which makes it difficult for them to pursue further education or secure good jobs.
According to survey, kids from this group tend to face 3 main life challenges.
1. Lack of money
The government tuition subsidy is not enough to pay for many necessary expenses
such as commuting, lunch, uniforms, and educational materials.
2. Personal problems
Some kids drop out because they have to help out with their family's work,
and some others leave school due to unintended pregnancy.
3. Lack of job goals
Those who finish lower secondary school are unaware of opportunities for further education
or what jobs the market demand.
Therefore, "Life Hero" was started...
in order to tackle these 3 challenges with integrated solutions.
1. Provide Scholarship
which will cover necessary expenses throughout three years of lower secondary school.
2. Provide Mentorship
Set up a 1-on-1 chat system to give kids continued guidance
that will help reduce the risk of dropping out.
3. Provide Counseling
for both career and further education through digital channels.
Last year, we were fortunate to have received enough support to help over 100 kids.
However, there are many other kids who still need help from heroes like you.
If you wish to contribute, please kindly make a donation of any amount at
www.lifeherothailand.com
Join us in a quest to create a new generation of heroes that will move Thailand forward!
-------------------------------------------
REALLY fast Russian – My education and why I teach Russian - Duration: 5:17.
Hello everyone! Daria here, and welcome to another fast Russian listening lesson
Today I'm going to tell you a little bit about where I studied and how I started to teach the Russian language
It will be in fast Russian, in normal pace
If you don't understand something, you can always use Russian or English subtitles
That's why I spend hours writing them for you
so please feel free to use them if you don't understand something
Ok, let's begin
Today I'll tell a little bit about my education
Where did I study? For how long? What did I study?
I'll begin with a school
For 10 years I studied in gymnasium in the city of Omsk
it's in Siberia
I liked my school very much
We had a lot of different subjects
For example, music, history of culture, choreography
it's about dancing
and even typewriting (where you learn how to type fast)
I loved my school
And after the 11th grade I went to the university
To St. Petersburg, my most favorite city, I adore it
I went to the University of Culture and Art
and began my studies to become a guide interpreter with English and German languages
And there I studied for 2 years
I was learning German, English, and I liked it very much
Visited different museums, galleries,
organised excursions for foreigners
And there, in Petersburg I started to teach Russian as a foreign language
There were students from China, and we helped them to learn, to get used to the Russian language
Later I went to Moscow, to the Moscow University of Culture and Art
to study culture and history of Russia
Also we studied the Russian language, literature, art – everything
And I graduated in 2011
Long ago...
After the university I went to postgraduate studies
and began writing a dissertation in Russian history – about the Russo-Japanese war and propaganda
I liked it very much
But the whole time I was still teaching Russian,
so I thought "I should go and study this officially and to get a certificate"
so in parallel with my post-graduate course I went to the Moscow State University
and studied to become a teacher of Russian as foreign language
After that I defended my dissertation, and got diploma and certificate
And now it turns that I have two professions
a history tutor at the university
and the Russian language tutor at the university
and that's my education
I like to study
and I hope to study somewhere else again
That was it. It was pretty long, because I can speak about education and about studying for hours
I'm such a nerd, so don't be surprised
And if you have any questions, please leave them below this video
And also I'll be happy to read about your education
where do you study? what do you like? what do you not like?
So share it with me
If you can do it in Russian, it's perfect. Try it
So... what else? I guess that's it
Thank you very much for watching. I hope to see you in my next video. Bye-bye!
-------------------------------------------
Education Bill Would Let Schools Expel Queer Students - Duration: 3:31.
The Trump administration is blocking an endorsement of international marriage equality.
Indonesian officials are intensifying attacks on queer citizens.
And a new education bill would let schools expel students in same-sex relationships.
We'll have the week's top LGBT news plus actions you can take on Weekly Debrief.
Hi, I'm Matt Baume.
Thanks to everyone who makes these weekly updates possible by pledging a dollar or more
per month on Patreon.
Now here's this week's news.
The US State Department is refusing to support an international statement that same-sex relationships
are entitled to legal recognition.
The statement follows a ruling from the Organization of American States, affirming the rights of
same-sex couples.
It's been endorsed by numerous countries in North and South America, leaving the US as
the sole holdout in the hemisphere.
Even countries like Chile -- which doesn't have marriage equality yet -- expressed support
for the ruling.
International support is sorely needed right now as South American countries grapple with
LGBTQ-related legislation.
For example, in El Salvador, lawmakers have been pushing a bill to block marriage equality
and adoption by same-sex parents.
But the country's Supreme Court just put a hold on that legislation, due to it being
improperly rushed through the legislature.
What's needed now is strong support for this international statement on marriage equality,
and that is in fact happening... but with one country's notable absence.
Meanwhile, there's a coordinated campaign targeting LGBT people in Indonesia.
Government officials have pressured Google into removing apps that serve the LGBT community
from the app store.
Indonesian legislators are considering a bill that would criminalize homosexuality, and
this weekend security forces rounded up and abused a group of trans women.
In the past, the US State Department would exert pressure on countries that abuse the
human rights of queer people, but now there's no indication that the US will lend any support
in Indonesia.
Australia just legalized the freedom to marry, but there's a worrisome backlash.
Anti-gay politicians have convened an inquiry into giving legal protection to private citizens
and businesses that want to discriminate against same-sex couples.
That inquiry is being led by Philip Roddock, the Member of Parliament who pushed the country's
marriage ban in 2004.
If you're in Australia, visit just.equal to sign a petition to ensure that all marriages
are treated equally under the law.
And in the United States, a bill is working its way through Congress that would be a dream
come true for schools that want to deny access to LGBT students.
The Higher Education Act establishes a wide range of government policies for colleges
and universities.
One provision in the current draft would let schools expel students in same-sex relationships;
or deny housing, aid, or leadership roles to students believed to be queer.
In decades past, the government could impose consequences for discriminating.
For example, the IRS threatened to revoke the tax-exempt status of Bob Jones University
for its ban on interracial dating.
This new bill would prevent the government from taking similar action to protect LGBTQ
students.
The Higher Education Act already passed a committee and is heading to a House vote.
But there's still time to stop it.
This week's action item is to call your members of Congress and tell them that the Higher
Education Act must protect queer students from discrimination.
A public outcry could prompt lawmakers to remove the anti-LGBT provisions from the bill.
Thanks for watching.
If you find these weekly updates useful, you can help support them by pledging as little
as a dollar a month on Patreon.
There's a link in the description -- just visit Patreon.com/mattbaume.
Or if you're not able to support on Patreon, just keep sharing these videos every week.
Let me know about stories that need coverage @mattbaume on Twitter.
And I'll debrief you next week.
-------------------------------------------
Vitalik Buterin Talks Blockchain Education - Duration: 6:23.
- All right, what's going on, guys?
Welcome to BlockGeek Youtube channel.
Today we have a very special guest with us,
Vitalik Buterin, the creator of Ethereum.
- Hello.
- Uh, so yeah, how's it going, welcome.
- All good, thank you.
Good to be back.
- So we just have some couple quick questions
on your thoughts about education and how a developer
can get started in this space.
So if someone were a developer,
and he wanted to get into this blockchain space,
what are some of your tips and advice to get started?
- I would say the best way to kind of really get
into this space is to
actually try writing your own application.
So think of some simple use case,
even you know just to start off
something totally useless like a button
that basically where if you click the button
it sends a transaction into the blockchain
in increments a number in some smart contract,
like basically see if you can do something
extremely simple like that.
You build it out the smart contract,
publish the smart contract,
build out the interface, publish it,
put it on a webpage,
see if you can use it,
see if everything works,
and once you're done with that,
then you can start moving toward more advanced stuff.
You know, as far as things that you should read
or look into while you're doing that,
I would recommend probably
a combination of various kinds of tutorials,
like there are some on our website,
there's some on other places as well,
I mean, I'm sure you have quite a few.
And, also, look at actual existing code examples,
and see if there's anything you can learn from them.
- Cool, cool, so it's like a very hands-on approach
versus reading whitepapers and things like that.
- Reading whitepapers is pointless I think.
(laughter)
Yeah.
- Okay, cool.
So I guess in terms of which area,
you talk about building applications but like
in terms of the Ethereum landscape right now,
or blockchain overall,
which area do you think lacks the most resources
where developers are most needed?
So in terms of scalability, infrastructure or building apps,
which area do you think lacks most resources
in terms of developers?
- I would say there's a lot of need for developers
even at the core protocol layer.
So, things like improving protocol scalability,
working on the client implementations,
implementations of things like proof of stake.
Layer 2 systems, things like state channels
and Plasma could probably use a lot more smart people.
On the application stuff, I mean,
I guess you could always use more,
but there's probably quite a lot more of that already.
- Cool, cool.
So I guess your thoughts about the education system,
like traditional education system,
like going through university
versus doing some kind of online training courses.
What are your thoughts about the different systems?
I know you kind of dropped out of Waterloo,
and went and did your own thing, so.
- I think different systems work for different people,
and anyone who's trying to force any single system
on everyone is probably trying to trick you.
Yeah.
The crypto space, in particular,
I'd say, and especially at this stage,
most universities are not going to have
that much that's useful.
Like, you can learn basic programming,
which will be useful,
but you can also learn basic programming
in a bunch of other ways as well.
And all of the crypto-specific stuff
is mainly available just on the online courses.
There are a few universities
that are starting off their crypto programs,
those are good as a starting point,
but for the majority of your learning
you probably want to be reading stuff on the internet
and doing things by yourself.
It'll probably continue being that way for a while.
- Cool, cool.
So I guess for someone who's watching right now,
and they're, you know, in a similar situation,
where they're thinking about dropping out of school,
starting their own thing,
can you walk through your experiences?
It must be a tough decision that you made,
and how did you come to the right conclusions?
- Yeah, and I think the main thing
that I was scared of when I was dropping out
was basically thinking,
"Oh, if I leave right now, what if I end up wasting
"the next one to two years of my life,
"and then I'll be far behind all of my friends
"who stayed at university and they'll know
"all sorts of various advanced computer science
"and I just won't have any of the background
"that everyone expects me to have."
That probably ended up being not true at all.
It might be different program-to-program,
but I personally found the value of university
is actually very high in the first year
but tends to drop off quickly after the first.
- Right.
- And in reality, if you're in the streets,
or in the field, and you're actually doing stuff,
and participating in various projects,
you'll end up learning a huge amount
on the way just like that.
So you'll be behind in some areas,
but you'll also be ahead in other areas,
and the ease of just learning any specific
thing you want to learn about on the internet
is definitely going up and up,
over the last few years especially.
So it's definitely not as hard as you might think.
But then, at the same time,
staying in university might also be
not as bad as you might think.
It's probably the sort of thing that you should do
if you really feel like you have something better to do,
and if you do then you should go for it,
and if you don't, then just stay there
until you figure something out.
I think that's totally fine.
- Cool, awesome advice.
So in conclusion, I guess you can talk a little bit about
what your future plans are
and what you're going to be doing.
What you're working on, stuff like that.
- I'm mainly focusing on the theory
of base protocol research, so Casper, Plasma, Sharding,
various scalability solutions, some privacy stuff.
- Cool, awesome.
Thank you so much, and really looking forward
to the Casper protocol coming out this year.
- Thank you.
- Alright, thank you.
-------------------------------------------
French Education System Explained: Grandes Ecoles vs University - Duration: 12:37.
Salut YouTube today I'm here with a real-life French person! Look, she's real!
Hi I'm Madeline, I'm from Paris. The reason I got Madeline into this video today is
because I've got this topic that I've wanted to cover for a long time which is
on the French education system so very much focused on the higher level
education we're talking bachelor's degrees master's degrees because the way
it works here is not obvious especially not to foreigners. There a big difference
between universities and what are called these Grandes Ecoles which are these
private schools which Madeline is going to tell us all about later so if you
want to know a little bit more about the French education system how it works
differently stick around and we're going to cover all of that in the next few minutes.
Madeline explain to us you know when you are at the end of your high
school years in France what options do you have?
Okay so mainly you will have like two main options the first one would be like
to go to what you said a Grande Ecole which is the equivalent of a private
university let's say most of the time and the other option will be to go to a
public university but what you have to know is that there is already a
different process in the selection like for public university you don't
have any selection so as long as you've passed it's okay most of the time
like there are some exceptions but let's say the big picture would be that
private Grande Ecole then you will have a higher
selection process most of the time also people do what we call classe
préparatoire which is like two years preparation to do the selection process
right so you study for two years just to enter a competition just to see which
school you can get into. Yes. Not for public universities but it's true for most of
Engineering schools, Business schools also. I've heard a lot about this "prépa" as they call it
and it sounds like hell. It sounds like very military like all my friends who have
done it had the same experience like they worked all day long all night long
you really study study hard and you're not even sure to get it at the end.
Isn't it true that with the grading system as well it's like really really harsh?
Sure! Sometimes for Classe Préparatoire you can have your profs telling you
oh you got 4 out of 20 it would be like nice piece of work like you've done
it well and all your friends around you got 2! So you're like WOW I'm the best of the class! It's so harsh, it's never
good enough. And I don't even get why they do that it's probably
to push you, push your boundaries, find your limits and see how far you can
go and how far you can study. I have to admit like when I came to France I
noticed that this was a really big thing about the difference between
universities and Grandes Ecoles. I've worked mostly in let's say CAC40
companies like the big multinational French companies and stuff so they're
very selective but you know whenever I meet people for the first time I'm like 'Hi! Cool, so
what university did you do?' and I get this reaction like I'm not sure if you know how the
French education system works but I didn't actually go to university
and we go to what we call a Grande Ecole and so I noticed straight away that there
was a stigma around it. According to you what are the big
differences in the students that go to university versus Grande Ecole?
I would say first like what you said sounds totally true. So students in France they know that their
choice about schools universities whatever will kind of define themselves
because in the workplace what is the most important in France is your diploma
and the name of your school and not necessarily let's say your personality
your skills or whatever and your experiences not always like it's really
your diplmoa that is important so people will choose between these two big
options. Grande Ecole the main difference is that they will really allow you to
network a lot with your peers with professionals you do a lot of
internships you've got a lot of professional experiences that you do not
have if you go to public universities it's very much about like get the job
get the job get the job whereas I imagine universities a little
bit still more around learning getting the content and thinking for yourself
and knowing how to be critical thinker etcetera so I think maybe
there's more of a more practical element of Grande Ecole but I was quite shocked
because I work in HR and It's very much been like
take candidates from the top five schools and I was like but we could be
missing like the most motivated talented person that's gonna hit the ground
running and shake things up in the company and they're like you know no the
managers say that they have become either from HEC, ESCP or ESSEC
In France I've noticed that managers will actually define like I
only want someone from HEC. That's so funny because what we say in France it's like
"Ils sont de la maison", like they're from the same house and it's
like Harry Potter you're from Gryffindor and so managers are like if they've got two
people if there is one that is from the same school as they were
they'll definitely, and it's natural also, like they're definitely gonna prefer the one
that went to the same university or school. It's a little bit bad for
diversity. I know know but I feel like there is a let's say just a little bit of change
around that and like companies are trying a lot more to find other people
than just the top five schools.
In some companies the people who do the Grande
Ecole they start on a higher salary than university students for exactly the same job.
Grandes Ecoles, if I understand correctly, you've gone through very
competitive exams to get in so you usually surrounded by kind of
intellectual elite but it's also it's costly to go to some of these schools so
it's also sort of let's say an echelon of society and so you've got a really
good network, you've got lots of professional experiences so you're building a
professional network. And so Madeline, you tell me, because you've gone to
university yourself and you're working in a really big company
right now so what would you say were the benefits of studying at a university?
Well one main benefit of going to university would be that is cheap
because the state is paying for that so you don't have to put three thousand a
year it's like three hundred so definitely
anyone can do it and that's why you got so much diversity
at University and that's so great like you can meet so many different people
it's enriching and interesting that you cannot really have in Grandes Ecoles.
Grandes Ecoles everyone is kind of the same and definitely years passing by
everyone becomes the same. Yeah you can sometimes tell, you can go to a party and you'll be like
oh I know which school you went to, you don't even need to tell me. So with the Grandes Ecoles and Classe Prépa
it's very very difficult to get into the Grandes Ecoles but once you're in it's
relatively easy to pass and because we're not so obsessed with academic
transcripts in France you can just get 10 out of 20 your whole way through
doesn't matter you don't care about your grades and it's kind of like once you're
in all you have to do is pass and you've got that name on your CV and you're fine
but at university it's a bit harder isn't it because it's very self-directed
because there is no selection at the beginning, the selection will happen later
like years passing by you will have exams not everyone's gonna pass so
definitely there is a selection so you have to be really autonomous one
thing you get from University for sure, independence,
it's not gonna be easy all the time and you're gonna have to learn by
yourself you know you have to read you're gonna have to, you know, think a
little bit like different like the only thing you learn at the University is
that there is no truth and every time you learn something then you have to be
critical about it. Grandes Ecoles also you've got a lot of parties you got a lot of
let's say student life you don't really have at university that's also kind of
weakness is that you do you do not really network with other people or
whatever but what I really feel that is, what I really like from University
is that you get this kind of share and participatory culture like
helping each other there's not such a competitive spirit like I don't know
it's really different like the mindset of the people is really different.
When you're in a university do you even kind of imagine working for a company? No no
It's really like for me like when I did my application to a big company private
company you know everyone was like "Madeline, why would you do that?" like there's no point and then I had
an interview and they were like wow you're lucky you like consider yourself as
lucky but you're not gonna go through that and it's gonna stop right
now then I made it. Why did they think that you wouldn't get in? Because this is
really like something that is in everyone's mind that for example from
University every time you will hear like you've got two options. First you can
do like research or you can work for public companies but stop thinking about
private companies this is for Grandes Ecoles. Because you hear that all day long you
kind of start to really think that way and you got people from university they
won't have enough confidence to think 'Oh, I don't care about the system,
I can do it myself and do my way and apply for private companies' and so it's
a vicious cycle because this is stigma about university students that may be
too theoretical and they won't be able to work in the business world
and the University students who are willing to work talented smart but
they're like oh I'll never get in anyway so you don't apply and so you don't get
considered and yeah it's so frustrating. I would say like if you have to, is there a real difference.
My opinion is that yes there is a difference because those worlds are quite
closed and do not meet so many times like if you're
from University you can hang out with people from University and same for Grandes Ecoles
However, the thing is like anything else both can adapt and easily and
what's important is the people behind it's because like it's not a real
difference it's a representative one. Yeah coz
you don't learn different things I mean if you're
studying communications in university or in a Grande Ecole it's the same content
sure no it is it's definitely not as simple as that and luckily it's not yes
I think this system has a lot of benefits a lot of downfalls and I know
that sometimes people go to university because they love to learn and they love
to be self-directed and then they'll go to the business school just for the last
year or two, have the stamp on the CV, I find these kinds of measures quite extreme that
you've got a lot of people in France with two masters one from university one
from Grandes Ecoles and even within the Grandes Ecoles network it's like oh
you just did your masters at Grande Ecole. So you're not a real one! I find this
whole system super fascinating but thank you so much for shedding some light for
us I'd love to hear what you guys think please comment down below especially if
you're French, if you've been to Grande Ecole, if you've been to University
honestly it would be really really interesting for us to continue this
discussion down there but until the next video I'll see you guys next time thanks
for watching à bientôt !
-------------------------------------------
Refugee students separated by travel ban take part in education program - Duration: 0:51.
For more infomation >> Refugee students separated by travel ban take part in education program - Duration: 0:51. -------------------------------------------
Leon Goretzka Lifestyle , Net Worth, Salary, House, Cars , Awards, Education, Biography And Family - Duration: 3:33.
Please subscribe my channel
-------------------------------------------
What's Working: Three Education Superstars in One Family - Duration: 2:28.
For more infomation >> What's Working: Three Education Superstars in One Family - Duration: 2:28. -------------------------------------------
Education Cartoons for Children😉Words with STR for Kids First Grade English Grammar Video - Duration: 2:54.
Education Cartoons for Children😉Words with STR for Kids First Grade English Grammar Video
-------------------------------------------
Cristiano Ronaldo's Lifestyle ★Net Worth ★House ★Cars ★Income ★Education ★Biography★Girl-friend 2018 - Duration: 5:26.
Cristiano Ronaldo's Lifestyle
-------------------------------------------
Reveal, Don't Conceal: Rethinking Data Visualization & Statistics Education to Improve Transparency - Duration: 1:01:45.
so for today's talk I'm going to focus on the importance of rethinking the way
our approach to data visualization and statistics education for small sample
size datasets and investigators who work with those datasets and increasing
transparency so the real estate crisis sure many of you know has a lot of moving
parts so people are thinking and depositing they are thinking about
registering studies they're thinking about a lot
study designs power calculation and physical analysis so that it's where we
well I would argue that data visulaization is important because data visualization is really the foundation of our
(Inaudible)
often put the data supporting key findings for the paper
and so if authors are making poor decisions about the types of figures
that they use um if they're using figures that are transparent they but
then providing the important information that we need critically at that data
then it for use that data and get the most out of the data set in the
future so today I'm gonna focus on several different topics I'll start off
with a quick overview of the problems with our current practices for how we
present data in small studies and that includes the inappropriate use of bar
and line graphs to present small sample size continuous data I'll talk a little
bit about why is this happened so how did we get to a place where the type of
figures that basic scientists use as standard practice are things that our
statisticians would say really just aren't appropriate at all for the data
that we and I'll talk a little bit about education so that the new generation of
scientists comes in with better skills and better understanding and I'll spend
the last half of the talk focusing on solutions so I'll talk about how you can
improve your static graphics to eliminate some of the bar and line
graphs and use figures that are more revealing and informative I'll
demonstrate a couple of our free tools for creating interactive graphics for
scientific papers I will show a couple of our interactive graphics resources
but I've also sent out from my Twitter account a few days ago a list of free
online resources that you can use to create better graphics for your own
scientific publications so if you have prism or other expensive software that's
great you actually don't need it to do
anything that I'm going to show you today there are already tools out there
that will allow you to do these things and I'll close with just a little bit on
the importance of meta research as a tool for addressing some of these issues
see rigor and reproducibility and if there are problems you are concerned
about how you can use meta research on your own to move your own field towards
solutions so our work we typically focus on small sample size datasets so
investigators who are working with less than 15 independent observations per
group and in the era of big data I think it's really important just to spend a
minute talking about why these small datasets are so important and why we
should be thinking about them small sample size datasets are extremely
common in basic biomedical and biological sciences as well as in small
sample size human studies and translational science and these studies
should not be overlooked they're critically important because they
influence decisions about what agents go on to clinical trials and to further
study in further research the clinical trials process as you all know is
time-consuming it's expensive it requires a lot of resources and time
from investigators and it places a lot of burdens on patients and so that means
that we want to have as much information as we possibly can when we're choosing
which agents are going forward so that we can make the best use of resources
possible and make good decisions and in order to do that we need to have the
small sample size studies presented in a clear and transparent way the other
thing that I think is really important to mention is that in a lot of the NIH
writings on reproducibility and rigor they are focused on the preclinical
research which is by nature small sample size studies so clinical
trials have undergone a lot of changes in the last 20 years regarding how they
are reported designed and conducted to eliminate or mitigate some of the
problems with reproducibility rigour and those things are just starting to happen
in the preclinical sphere so that means preclinical research still has a lot of
problems and there's a lot of things that we can do to improve
reproducibility and rigor in these small sample size studies so from the NIH
perspective looking at the preclinical research is definitely somewhere we
should be focusing whenever we're thinking about reproducibility and rigor
okay most of us never get visualization training that we need in order to do our
scientific jobs so our visualization training is often we get our first data
set we turn to the grad student at the destiny next to us and we say what do I
do with this and they say you make a bar graph you make a line graph and and we
continue doing that for the rest of our careers this isn't a very good idea and
it's not a very good way to train people in visualization so today I'm going to
give you a little bit more information on how we can design better figures and
why that's important so an effective figure needs to do three things the
first thing is that it should immediately convey information about
your study design so I should within seconds of looking at your figure I
should know whether you're comparing a set of independent groups whether you
have longitudinal or repeated-measures data or whether your data set includes
clusters of non independent data so things like replicates or animals from
the same litter figures should give me all of that information the next thing
that the figure needs to do is illustrate the important findings for
your data set so again within a few seconds of looking at the figure I
should be able to get a sense of what it is that you're comparing or assessing
and what your key findings are going to be and then the third thing is really
critically important and that is that your figure needs to allow your reader
to critically evaluate your data this is what differentiates science from
marketing right so in science we need to be able to look at each other's data to
critically about to draw conclusions and then make
decisions about how we can best to move the study forward and advance knowledge
by conducting other studies if the data are presented in a way that don't allow
you to critically evaluate the data none of that process happens and that's a
huge limitation to our scientific method so this is the major limitation of the
types of figures that we use now the bar and line graphs that they don't allow
the reader to critically evaluate the data and just so we're clear as dr.
Heywood mentioned my main interest is in pre-eclampsia research and my interest
in improving the other presentation actually comes directly out of my work
in preeclampsia so with preeclampsia it's a syndrome and we know that well
women all end up with the same symptoms they tend to get there in different ways
so some women have a lot more maternal pathways that are contributing other
women have a lot more pathways related to the fetus and placenta and the
contribution of those maternal pathways and the fetal and placental pathway
differ from woman to woman so that means that for any marker that we're looking
at we expect that it's going to be abnormal in some women who get
preeclampsia but perfectly normal in other women who get preeclampsia and our
goal right now is the field is to understand how we can identify which
different subgroups and which women have normalities in which pathways and which
women have normalities and other pathways that are also relevant so if we
want to understand heterogeneity then you need to start presenting our data
differently a bar graph is what we use if we want to mask heterogeneity right
there are no subgroups in a bar graph so we started off two years ago with our
first meta research paper which was published in PLoS Biology and it was
titled beyond barn line grass time for a new data presentation paradigm and this
paper went viral online on Twitter and Facebook among the scientific community
within about 48 hours of publication and it was picked up by Nature News and the
Twitter response at that time as well and it was viewed more than a hundred
thousand times in the first month that it was published since the time that it
was published it has had a number of effects so it has led to policy changes
in several different journals including PLoS
Biology life and the Journal of biological chemistry that are now
discouraging or banning authors from using bar graphs or continuous data at
least for small sample size datasets nature also came out with a policy
change a few months ago that applies to all of
their affiliated journals as well we've also know anecdotally that a number of
editors and reviewers have been using the paper to request better data
visualization for papers that are going into different journals and it was also
instrumental in initiating the bar bar plots Kickstarter campaign which
targeted neuroscience journals and general circulation journals to improve
data visualization policy and get rid of some of the bar graphs so the paper had
three different parts we had a systematic review that looked at
quantifying how do we present our data now and what are the data presentation
practices we had a clear explanation of why our current data presentation
practices are problematic and then we also had templates that investigators
could use to make dot plots and better graphics that were done in Excel so we
looked at 700 papers that were published in the top 25% of physiology journals
over a three-month period and we found that almost all papers were using bar
and line graphs to present continuous data so 85% of papers had at least one
bar graph of a continuous data variable 61% had a line graph when we looked at
papers that were using figures that showed the data distribution those
papers were rare so only 15% of papers had a dot plot and
five to eight percent had a histogram or a box plot so when it comes to figures
that showed a data distribution in the basic sciences we tend not to use those
and that's a problem we also found that the sample sizes for very small and in
most cases less than 10 per group so when we looked at the minimum sample
size for any group that was shown in a figure and 75% of papers that sample
size was 6 or less when we looked at the maximum sample size for any group shown
in a figure in 75% of papers that was 15 independent observations or less so
these are very small data sets we also found that investigators who are using
bar graphs are predominantly using them to show the standard error
which tells you about the accuracy of the mean and not the variability in the
data so 78% of papers with bar graphs for using those graphs to show the
standard error only about 16% we're showing the standard deviation and
finally we found that these data presentation practices were so ingrained
that even amongst papers that use some form of nonparametric analyses more than
half of authors still chose to present data that were analyzed non
non-parametrically as mean and standard error or mean and standard deviation and
those summary statistics aren't appropriate when you're using a
nonparametric tests so we use a lot of bar graphs who cares really and why is
that a problem well one of the reasons it's a problem
is many different data distributions can give you exactly the same bar graph and
your actual conclusions may be different based on the actual data set so here we
have a bar graph and then on the right we have three to four four different dot
plots of data sets that will give you exactly that same bar graph and there
are p values for several different types of t-test below so t-test for equal and
unequal variances as well as a nonparametric wilcoxon test so if we
look at our first dataset we seem to have slightly higher values in our
second group compared to our first group and this might be this is a small
difference but it might be one that we're interested in pursuing when we
look at our second graph the higher values in the second group really seem
to be driven by this one outlier and that's something that we might might
less might be much less likely to follow up on when we look at our third group
there's some suggestion that the data distribution might be my modal so we
would want a larger just the larger data set in order to confirm whether or not
the distribution is in fact bimodal and we'd also want to know is there
something different about these observations compared to these so are
these the males and these the females is there some other variable in our data
set that we should be accounting for and then in our last
should've unequal end so our data police in the second group are clustered to the
top end of our range for the first group however we have far fewer data points so
it's entirely possible that we have just underestimated the variability and if we
had more data points in here we might see much less of a difference between
these two groups so we will probably want more data points within that second
group before we went towards any kind of conclusion with that data so the next
question that I always get is okay well let's say I know my data are normally
distributed can I use a bar graph then and the answer to that question is still
no for a couple of reasons first it still is not going to allow you to
critically evaluate that data so you've said I know my data are normally
distributed but this is science and we want to see the evidence of that we want
to see the data showing that so though we can say yes we agree with your
conclusion that's correct and that's how we would have done it too. the second thing
is the bar graph actually draws our attention to all of the wrong things in
(Inaudible)
at the end of our bar graph is typically going to be just of the highest group or
the the highest error bar for the highest group shown in the graph and
when we do that we actually omit values from the graph that are actually
observed
(Inaudible)
zone of invisibility these are these high values here that occur in the data
set but don't actually show up in the graph and then we have a second
problematic zone here now when we make a bar graph we typically start the y-axis
at zero in some cases zero is a physiologically or biologically
meaningful value in many cases it's not variables where zero is simply not
possible it's something that's never going to occur and so we can end up with
this low set of values on our graph that aren't a value that we would ever
actually observe in our data set they have no physiological or biological
meaning and we call this the zone of irrelevance so when you make a bar graph
what you're drawing attention to is you're missing the zone of invisibility
you're putting a lot of weight on the zone of irrelevance and you're
arbitrarily assigning importance to the bar height instead of focusing your
readers attention on how the difference in means compares to the variability in
the data or how much overlap there is between groups so just to emphasize our
interpretation depends on what we see so with the bar graph all I can really look
at is do these groups look different on average there's no information here
about the sample size that might tell me how confident I can be in these
conclusions there's no information about the data distribution there's really
nothing that allows me to critically evaluate the data in contrast and that
turns the reader into a passive observer in contrast when I have a dot
plot here showing the data points I immediately get a lot more information
that causes me to actively think about the data set and about the conclusions
that are being offered so for example I might say okay and I see is that this
small sample size data set so I should be cautious about these conclusions
there's a lot of uncertainty here I might say this group could potentially
be my modal and I might want more observations to check that I would see
that I have a pretty clear outlier here I might also see that while this group
has high values it's also the smallest group and we've probably underestimated
variability there so again there's a lot of uncertainty into whether this high
bar is really meaningful and different from the other
groups so in conclusion all of the data
presentation methods that we use are a reflection of reality and it's really
important to use methods that minimize distortion looking at your data in a bar
and line graph is kind of like looking wavy pond maybe you're looking for a
duck and you see it and you think that it's a duck but before you waste a lot
of time and effort in following up on those findings and pursuing something
you want to be confident that that thing that you're seeing is actually a duck
and not just an oddly shaped potato right nobody wants to waste money
looking for an oddly shaped potato when they thought they were chasing down a
duck okay so why does this happen so to give you an idea between the this
community and the stats community I have two statisticians that I work with and
when I went to them and said you know the problem that I'm worried about
scientists use bar and line graphs all the time here's my data and I'd like to
do something about this one of them didn't understand what I meant I said
you know she said to me like how can you use a bar in line graph for continuous
data I don't know what that means and so I had to explain the bar is the mean and
then the error bar is usually the standard error the standard deviation
and she was very confused about why anyone would do this and it took me 10
to 15 minutes to explain the other one said I don't believe you and I said but
I have you can't not believe me and she said I
understand like I see your data and if that's correct that is absolutely
horrifying I just can't believe that anyone much less a field of people will
be making a mistake that's this basic and I said well if you're following
standard practice for your field you're never going to ask be asked to justify
or explain that practice you're just doing what everyone else is doing and
everyone accepts that's the way you do it no we get here how do we get to a
place where the practices that we follow our basic science as basic scientists
are things that people with more training in statistics and in data
visualization would say are completely inappropriate we in a second paper that
we published in PLoS Biology and one of the things we found was that
while statistics are essential for anyone who is reading papers or
publishing in the basic scientists statistics training is not always
required for a basic sciences PhD so we started off with our data set again of
700 papers published in the top 25% of physiology journals and we found that
97% of them included some statistical analysis which is probably not
surprising to any of you who picked up the dirt like although in surprise there
it was we looked at the top nih-funded physiology department and there were 80
of them and we looked at whether or not students were required to take a
statistics course for the PhD programs that those departments were
participating in and we found that in about 2/3 of the cases that statistics
was required for some or all PhD programs of the department participated
in the remaining third it wasn't required and there were a mixture of
different strategies so about 10% included stats in their program as a
recommended elective 10% included it elective that but didn't specifically
recommend encouraging students to take it and for the remaining 12 and a half
percent statistics was not required it was not an elective it simply was not
part of the program and the paper is among our 80 departments we had about
five that had just changed to requiring a year that we did the survey so if you
think about people and five years ago suspect these numbers would look
substantial than they do now the other thing we found is that even when
students are taking there just because this is four that's
and a lot of physiology students who were taking courses with titles like
statistics for public health which is large sample size data and it's
completely different a lot of departments don't have statistical
expertise in health or in house so they will go genealogy or bio stats or
another department Public Health psychology an apartment that has more
statistical expertise to get stats training for their students and the
result of that is you often have director who is used to working with
large sample size data and assumes that their students are going to be doing the
same thing teach those students who work with small sample size data sets with
shins and study designs and neither realizes what's going on so if you look
at the statistical problems that we have in the basic science literature from the
perspective of instructors teaching assuming that students are going to be
working with large sample size datasets students hear it assuming that it
applies to the small sample size with practices accordingly then the problems
that we see make a lot more sense so we firmly believe that basic
scientists really need to be actively involved in designing stats curriculum
and in making sure that instructors understand what questions their students
are asking what techniques they're used to seeing in the literature for stats
what the problems are with the way stats are typically done what sample sizes the
students are used to and both statisticians that I worked with when
you know for the when they started reading basic sciences
paper as they said okay if I had known that this is what my students were doing
I would have behind my courses completely differently to begin with and
so at Mayo Clinic that's one of the things that we're doing now we are
developing new curriculum for our basic sciences at Mayo and once we have that
curriculum it will also be available online for anyone else at any career
stage who wants to use it so we do know that are different when we're planning
our courses the first thing is focus on data visualization first second so
teaching data visualization is a lot easier it's a lot less intimidating for
many students and when students start seeing their data it needs it naturally
leads towards a whole range of statistical questions it creative piques
their interest and it makes it much easier to move it to the statistical
topics naturally the next thing we're doing is targeting misconceptions and
missed skills that are common in the basic sciences so misconceptions would
be things like you should present continuous data using a bar graph or if
you find a significant and effective than a small sample it must be a really
big effect or it doesn't matter if your study is underpowered as long as you
find a significant effect so these types of things are all incorrect but they're
all things that many basic sciences have been taught to believe and we address
those things head-on with our students because we know we're gonna they're
going to go out into a world where these practices are common so if they haven't
heard them and they don't know what the problems are it's going to be very tough
to make any kind of lasting cultural change the myth skills would include
things like analysis of clusters of non independent data so replicates or mice
from the same litter basic scientists worked with these type of datasets all
the time we're never taught how to analyze them how to tell whether your
cluster design is between group clustered within group clustered or
between within group clustered why that matters and how it affects because every
one of those things requires different sets so we teach those skills upfront so
that we know our students have them or at least aware of the issue we also use
the visual approach to learning so we'll create a set of visualizations
and a lot of maths we'll start with the
visualization and use that to have a conversation about the issue and to
introduce the topic and we're also creating simulation tools to allow
students to really play with some of the concepts and the rules that they get
their statisticians so that they can understand what what considerations and
thinking are going into the that they finally get and how arbitrary some of
those rules actually are because the thing that's been striking for me is how
much of what I was taught and basis in my basic statistics training that I have
had to relearn as a result of going through this process of critiquing the
literature so here's an example of the type of visualization that we would use
in one of our simulators to teach basic concepts so here we've created a data
set that has a mean of this black line and then plus or minus one standard
deviation is shown by the gray shaded region and then we just draw random
samples from that distribution and we did it twice once with an N of 5%
sample and once with an N of 20% and we're just looking at how much error
there is in the summary statistics and how that depends on your sample size so
right away that you can see that if you have an N of 5 per group there's a lot
more error in your estimated summary statistics they're much less accurate so
the means are further off in the N of 5 compared to the N of 20 the standard
deviations are also further off and the problem is that we will do your one
experiment with your N of 5 maybe you have this sample here
honey standard deviation or maybe you have this sample all the way out here
that's way off or the sample over here that's also way off but in a completely
different direction if you're working with a small sample you're going to have
less accuracy in your estimate and there's no way to know which of these
specific samples from this or any other possible sample you might have drawn I
don't know how far off you're going to be so this is the type of visualization
that we use to show that with small sample sizes means and standard
deviations can be very inaccurate we then go to something like this which
helps to quantify okay you're it so here's our n of five and here's our n of
twenty and we're just looking at the error in the sample mean here so this is
no error at all what it should have been in the population if we're up here then
the mean is a half a standard deviation off here or sorry quarter standard
deviation off half a standard deviation off three quarters and standard
deviation off one standard deviation off so the higher you are the more error you
have and we can look at the end of five we can see that small error so less than
a half a standard error less than a quarter of a standard deviation for N
of 5% per sample only 43% of our samples are meeting that criteria but if we go
up to an N of 27% percent of our samples are meeting that
criteria all right so this helps students to
think about what do I get what's the benefit if I go
to a n of 15 out of 20 or 9 of 30 and is that possible for my experiment this
is another example of a simulator that we've created to talk about the issue of
normality so many of us in our intro stats course were given some kind of
role like where a sample size is smaller than 10 to 30 you should just go
straight to the normal test because you can't tell whether your data are
normally distributed or not what you get depends on and you had I've heard people
give in 10 and 35 for those values so ok does that number come from
who made that up and why was it that number and if that were so accurate
every statistician number later is just health questions so we start off with a
visual simulation and we say okay just using your eyes
how your data are distributed normal distribution is skewed distribution and
a bimodal distribution and they can enter three different ends and then they
can click draw a new sample at any time to just generate
samples so if n is 100 I think we can all be pretty confident that this looks
normal this looks skewed and this looks bimodal
and most of us would have no problem there when we get down to an N of 20
life gets a little bit more complicated this we could say looks probably normal
ish this may or may not look skewed might also look slightly normal hard to
say this does look rather bimodal but we're starting to see some uncertainty
here and how confident we can be when we get down to n of 5 there's very little
information and they all look the same nobody has any idea what's going on we
can't tell so we'll ask our students ok what happens if you have an end of 100
what happens if you have an N of 5 and then if you had to pick a number saying
where can I wear my confidence that I can tell what number would I pick and
they can play with all kinds of different numbers and figure out what
answer they would get we then come and say ok was visually with your eye what
happens if you do a normality test how does their morality tests work so we
have their same three distributions normal skewed and bimodal and first off
we need to talk about what we're going to expect so if the data are coming from
a distribution that's actually normal we're gonna expect that 5% of samples
are going to fail that normality test right this is a type of 5% are going to
be extreme in some way so even though they came from it as normal distribution
mostly identify them as being not normally distributed on the other hand
if our sample came from a skewed or a bimodal distribution it's definitely not
normal so we would expect 100% of those samples to fail a normality test so what
actually happened well here we have the same setup so we can enter our sample
size so we have 120 and 5 and if we have a hundred then it looks about like it
should we get 6% with the normal distribution and 100% with this
distribution so just like our eye, the normality test as well when we go down to an N of
20 our normal distribution we're still getting about a 5 percent failure rate
but we start to see something very different happening here for the skewed
and bimodal distribution now in this case more than around half of the
samples that are from those distributions are still
filling our normality or are still passing our normality test so the test
is having a hard time distinguishing just like we did by eye as to whether or
not data came from a normal distribution when we're down to an N of five we have
a ten percent and a twelve percent failure rate instead of 100% failure
rate so just like our eye has no idea also
the normality test has no idea and again students can enter any numbers that they
want in here and get a sense of what do they get additional five samples yeah
I'm not sure
but we can also change it and be like this is all done so it can be changed to
be whatever normality test we want to use yeah oh the question was what kind
of normality test was used okay so in terms of solutions I'll talk about
a couple of things for static graphics and alternatives to bar and line
graphs really send things out from my Twitter feed as well that's another you
can look for information and I'll probably send out a bunch of stuff
tomorrow following this talk okay so if we're not using bar graphs what are we
using instead and how do we decide because there are multiple different
options use well it depends it depends on your sample size it depends a little
bit on your data distribution and it depends on the goal of your graph so
assuming that we have a continuous outcome variable and a small sample size
so less than my arbitrary number is going to be 15 it's an arbitrary number
telling you that full disclosure so in this case you are going to use a dot
plot Y as we've just seen with the simulator summary statistics are only
meaningful when you have enough data to summarize when your datasets are really
small those summary statistics can you're it so it's best to simply show
this is my actual data these are the points this is the observations that I
had if you have a slightly larger dataset you now are going to get into
the point into the point where the summary statistics shown in a box plot
are going to have some meaning and your sample is large enough to actually
calculate these so in this case you might consider using a box plot with the
data points overlaid to allow readers to see not only your sample size and a
little bit more information about the data distribution but again to still be
able to critically evaluate the data if you have a large data set then it
doesn't make sense to show the dots you'll end up with a swarm of dots and
they'll be really hard to see anything so the summary statistics for your box
plot are going to be accurate and you can simply present a box plot critically
important thing here if your data distribution appears bimodal you should
not use a box plot ever a box will mask bimodal distribution so there
will be no way to see that in fact you have a peak here and a peak here that
information is going to be completely lost in the graph so bimodal
distribution never use a box plot the other alternative if you're not a fan of
box you know there are lots of basic scientists who aren't it's to go with
something called a violin plot and you can show the data points on this or you
can take the data points off depending again on your sample size so again
useful for continuous variables medium to large sample size and you can use
this with any distribution so it will show a bimodal distribution if you use
this type of plot and essentially what you're doing here is you're taking all
the data points
mr. B and the last thing is your bar graph
your bar graph you don't use for continuous data you're going to use it
for counts and proportions and it can be used for any sample size so a lot of
investigators have started to create box plots when we're creating effective box
plots it's really important that all of the data points are visible so I'm going
to walk you through a series of strategies that you can use to make sure
that all of your data points can be easily seen so we start off with our
ineffective graph here we have a lot of data points overlapping it's a little
bit hard the data are distributed and what's going on so the first thing that
we can do is decrease the size of the points this helps it cleans things up a
little bit but we still have kind of a mess with a lot of overlap so the next
thing we can do is make those data points semi-transparent when we do that
regions with overlapping points are going to come up darker than reasons
without the next do is random jitter so we add some noise on the x-axis and
again this is designed to reduce the overlap the problem with random jitter
is it never fully eliminates the overlap unless you have a really small sample
and it also isn't very effective for showing you the data distribution or
simple or a more visually appealing option and visually easy to interpret is
symmetric jitter where you try to line points up on either side of the x axis
or along the x axis so that they're symmetrically distributed and you can
get a sense for what the data distribution would be the next question
I'm trying to make these dot plot things but I have a lot of groups in my graph
and you know my figures should give a clear indication of what my differences
are and nobody can tell it with these dot plot things so I like to bar graphs
better so the first thing you're going to do is increase the width of your plot
white space
(Inaudible)
the next thing you're going to do is emphasize your summary statistics so
here I've shown my means and my arrow bars in black so
groups look different from other groups but then I can also back and look at the
data points which are in the background in gray thence for does everything look
like it's distributed completely for the summary statistics and the tests that
are used. I can evaluate the data set in more detail how do you work with figures
that are larger that are more complex that have a lot of different groups so
while our initial paper has been very effective in encouraging some journals
to state to change policies we've now moved to a different strategy for
encouraging investigators to provide more information about their data in
their data distribution so we publish a third paper in PLOS biology
from static to interactive transforming data visualization to improve
transparency and the objective for this line of real research is to really
create the tools that investigators need to transform their scientific
publications from static reports into interactive datasets narrated by the
author so if we think about the incoming generation of scientists and
many of you are in the room are young I'm sure you've all grown up with smart
phones and tablets and every interactive app you could ever want and the ability
to manipulate a lot of data very quickly this is the future of science if the
weather is interactive certainly science can do better and so we think that it's
unlikely that it's more and more of us gets used to using these technology and
tools we're going to continue to believe that the best way to present our data is
in a little tiny bar graph printer them black and white on a piece of paper so
we started off with the line graph we had a lot of people emailing us and say
you know I really like being able to see the individual level data
we dot plots that adds a lot of information but how do I apply those
same principles to a line graph and there are a couple of problems with the
line graphs there are two pieces of information that we really want to have
a scientist and they're hard to get out of line graph the first thing we want to
know is how much of different groups overlap and depending on the line graph
this can be really difficult to tell we often end up with something that looks
like this where we have a lot of overlapping error bars and it's hard to
tell which set of bars goes with which group the overlap it the second thing we
want to know is do all the individuals in a particular group follow the same
response pattern there is no information about that on a line graph a line graph
tells us what happens to the group on average but there are no individuals in
this graph so we get no information there so to address this we created a
free web-based tool anyone can access it by going to this website and it allows
investigators with no programming expertise to recreate an interactive
line graph for their next scientific paper so if you can or check your email
or check the weather online you can do this you can use this to make an
interactive line graph I'm not going to demonstrate it today but there is a set
of video instructions on the site it's links that you can just go a video of me
and how to use the different parts of the tool and I am gonna show you a bunch
of screenshots just to illustrate how easy this is and what kinds of things
that can do the interactive line graph allows you to do or things first we can
examine different summary statistics second we can display lines for some are
all individuals in each group third we can view a subset conditions or time
points and for you've changed scores for any conditions or time point first step
is getting your data into the graph and there are two options for how to do this
if you are a manual data entry person you can enter your data manually so you
can tell it the number of groups and conditions give it some names for your
groups and your conditions and some sample sizes and then you can either
click on save and proceed to enter data and you pin your values or you can
download a template CSV file download a CSV that you can open in Excel it'll
give you all of the headings and all of the
and you just type in your numbers into that and you can then upload your work
set into the data file so an XML is for that we a CSV file is if you are
creating a new graph with a new set of data then you'd upload a CSV you just
select your data file upload it and hmm the program will automatically create
the graph for you so when your graph comes up you will see something that
looks like this and the first thing we've done is we replaced the error bars
with a semi-transparent shaded region so this is going to make it much easier to
see which groups overlap this is the normal mode
who has a colorblind mode so it's just one click if you're colorblind click
that and no that's much more easy for you to interpret the next thing the
thing you can do is a tool it selects any option for your summary statistics
so using show means or medians you can show means with standard deviation
standard errors with 95% confidence interval show a median with an inter quartile
range or a range so that allows you to pretty quickly quick click through and
see a lot of different summary statistics quickly the next thing you
can do is focus on specific groups time points or conditions of interest so if I
go back to my main graph it looks like there's a lot of variability in group to
the right of group compared to the other group and I might want to know why that
is so one of the ways I could get that information was to focus on group two so
if I unclick the boxes next to any group or condition it will remove those from
the graph and I can just add them by rechecking those boxes at any time so
here I've unchecked Group one and group three so now I can just focus on group 2
the next thing I can do is view lines for any set of individuals in the data
set so I can select all or none and this will show me the lines for everyone and
when I see that it looks like I have some people that are responding and some
people whose values aren't changing here in condition 2 and I can fiddle with
this in more detail to figure out which individuals are responding in which
aren't and to just be able to look at the responders the last thing I can do
in order to confirm is I can go to the difference plot tab and this allows me
to create a difference plot for any set of time points or conditions so I can
change whether I'm looking at the mean or the median for the line here I can
select condition one to two but I can also do two to three or condition one to
three anything to conditions that I want and
then I can also remove groups in the graph if I want to focus on certain
groups and when I look at the data here it does look like I have some non
responders and some responders here in group two which I don't see in the other
groups the interactive tool also has some features for linking to
publications and in particular we want to make sure that the static graphs that
we present in our paper aligned with the interactive graph
that we are giving people to play with so there are some functions that allow
you to save any static graph that you create using the tool in your paper so I
can simply enter a name for the figure so here I have figure 1d for my paper I
click Save once I do that the figure 1d is going to show up in the build menu
and I can put as many figures in here as I want and this is a useful tool because
it allows you or your reader to explore your entire data set or just to say I'm
interested in figure 1d I want more information on that so I'm just going to
go directly to that and start my exploration from that figure and then
the next thing you can do is download so there's a several different ways of
downloading and there are clear and easy options on the file I can download a CSV
data file if I just want to get the data out and run some stat I can download
TIFF files for any of the stave static grass and this is what's going to go
into the printed version of your paper and then I can also download an XML file
containing your interactive graph like to include in a supplement for the paper
and that allows (inaudible) XML file back into the site to
explore the data in more detail if they'd like to do that important thing
to know when using the site we do not save
I don't care so it's important to save your work and save your XML file if
you're going to want to go back and make changes to that graph later because we
know how to save it before recently we released a tool called the interactive
dot plot so this is an alternative to the bar graph it came out at the end of
December for those of you who are wondering about the title of the paper
we submitted it with a much more mundane title and the editors come asked if we
could come up with something that would be more fun and a play on words
so we obliged so now I have a scientific paper location that includes the word
naked in the title all right so the interactive dot plot tool will allow you
to make a whole range of different types of graphs and which one you're going to
want to choose depends on your sample size and the other stuff we talked about
so you can look at a dot plot you can add arrow bars to that job pod or a box
plot to that Docs to that dot plot and there are options for putting the docs
plot dot plot in the front and the back or vice versa you can also do violen
plots and this is because a lot of commercial software packages don't to
make file in plots so here is a convenient free and easy way of doing it
for those of you who want to use those you can have a dot plot with data points
on it or a violin plot with data points on it or within box plot in the middle
and we will actually even allow you to make bar graphs but if you make a bar
graph without data points message that will say the bar graph has
included as a teaching tool and is not intended for use in scientific
publications the relevant publications and initiatives and those numbers will
take you to links that it would will explain why it's a terrible idea for you
to use a bar graph for your paper you can do it if you want
a tool has a couple of other features you can examine suburbs so for example
if you're working with a data set that includes males and females you can
actually allow your viewer to look at the data for the males and females
separately so you can color code the males and females you can choose to show
them side by side or you can just show them all pull together in one group and
you can add summary lines to show the median or the mean for your males and
females and I could also uncheck these to show only the females or only the
males if I wanted to do that the next feature it allows is for you to show
clusters of non independent data so this is your replicates or your mice from the
same litter and it will automatically do a between group cluster design so this
is where I have one observation from each cluster in each group it will do a
within group cluster design so this is where all of my observations from
cluster 1 and cluster 2 are in group 1 and all of my observations from clusters
3 & 4 are in group 2 and it will allow you to do a between a group within group
cluster design so I have observations from each cluster in each group ok so
when we're thinking about our interactive tools we envision two sets
of solutions one is the author level solution so these types of tools that
I've shown you that allow you as or as to go in and quickly create an
interactive graph for your paper the second one is journal level solutions so
we're very interested in working with online graphics a part of the scientific
publication so load some data and then upload it into
our site you can just play with it directly on the journal site as your my
version of that paper and we'd also like to make sure that those solutions are
compatible so the stuff that you create using our author level solutions we hope
at some point will just be able to automatically be fed into the journal
website so that the journal can immediately use that and you won't have
to create another whole set of stuff when you're publishing because we all
hate that okay so I just want to emphasize that changing present data
presentation for continuous variables in small sample size studies is really
critically critical to promote transparency so right now we have the
bar and line graphs and then we have everything else and we would like to
switch this around so we would like this to be figures that show the data
distribution that are clear and transparent and this is the bar on line
graphs and I would appreciate the help of everyone in this room and continuing
to move that process forward so here are some things you can do to help the first
and obvious thing is to banish bar graphs from your papers and talks and if
you are a reviewer or an editor for a journal and I would encourage you to
request figures that show the data distribution as a standard part of all
of your reviews it's also important to talk to editors for journals that you
read regularly about the importance of improving data presentation in their
journals and offering more transparent options and lastly I think it's
important for investigators at any level to be working with statistics and
presentation and statistics training and things that would be appropriate
not just for trainings but also for junior investigators as well as senior
researchers and lastly I want to emphasize that it's important to be a
part of the solution so I plan to have a partial career in Metairie search that
was something that happened accidentally because I was frustrated by something
that was going on that I felt was limiting my fields ability to move
forward and I'm sure that many of you in this room have similar concerns about
other practices be they in stats or in methods or any other area of your choice
so it's always a possibility to read meta research studies to learn about
existing problems discover better prac identify problems in your field and see
how fields that may be further along in this area have problems and whether
there are solutions to your own fields and into your own work and finally for
those of you who have particular issues that you're passionate about and
concerned about I would encourage you to start talking to people and to bring
together things and perform your own meta research studies to solve a problem
that's of importance to you in your field all right I would like to thank my
collaborators some of the mentor resident Garovic at Mayo the to status
students I work with our Stacey Winham and Natasa Milic who's the head of bio
stats at the University of Belgrade Medical School Ethan does all of our
simulators and Marko Pro does all the programming for our interactive tools
that I'm ready to take lappers okay and so the question is how do we start
addressing these concepts to do statistics education earlier before students get to
grad school and how do we present them in a way that resonates with students we
don't have an undergraduate program at Mayo so there's not a whole lot we can
do there so we are essentially dealing with graduate early career investigators
who are looking for information and sometimes pis
so we do do a lot of we'll do a particular lab wants a seminar on how to
present data or an instructional session at lab meeting we will do a lot of
outreach and things that way to get information to lab groups that are
interested I think what you said about targeting examples and setting up things
in a way that's appropriate for the audience and will resonate that with
them is critically important and that is the centerpiece of everything that we do
so every time we work with the new statistician I give them a full one-page
10-point font no spaces list of everything that we know about our
audience so we talked about what study designs they're using their level of
statistics education what software they're using what mistakes are common
in the field what sample sizes do they work with and then we have a full
conversation about how the particular problem that
we're interested is in is going to manifest in the published literature for
people in that field so we make sure that our examples are just either one of
two things either discipline appropriate wherever possible or general enough that
people could see their own data in what we presented so things like some of the
figures that I showed they're very vague but I think many of
you in the room who are basic scientists could look at it and say okay yeah I've
seen data that looks like this I get data that looks like this so
whenever we are writing a paper or doing anything we spend other than data
collection for a meta research we spend about 30 percent of the time on content
and 70 percent is on our communication plan and how are we going to present
this in a way that resonates with our basic science audience so I would say if
you're not already working with a basic scientist or two or three find one or
two or three because those are the people that can really help you in terms
of saying you know because you're as a statistician you have a different
mindset you have a different way of thinking so
the argument that's going to work with a roomful of statisticians is clearly
they're going to be the argument that's going to work with a roomful of basic
scientists so having a basic scientist that you collaborate with closely I
think is really essential in making sure you're getting the right information to
the right people and at the right many things are possible seeing the data is
really important and it's important for a lot of reasons so the question was
about papers that have bar graphs with a lot of groups that overlap and arrow got
the arrow bars also overlap but the differences are significant so I think
one of the things that's critical and that's a problem with basic sciences is
a lot of basic scientists aren't aware of what the arrow bars represent so
really checking is it standard deviation versus is it standard error because with
standard deviation you will get a lot more overlap even if there is
significant difference so that would be the first thing I would look at and then
I think it's also really important to think about statistics was the technique
that they're using appropriate for the data that they were
analyzing and so we've just finished a study on reporting for ANOVA where we we
were asking the question among papers published in top physiology journals
that we're using ANOVA can we tell what type of ANOVA that they did do we have
enough information to verify the test result and we also looked at the
prevalence of two common errors so papers using a one-way ANOVA when two or
more factors were needed and papers that were not using repeated measures or
didn't say they used repeated measures when there was a variable that clearly
seemed to require repeated measures and the results were horrifying and it's so
it's surprisingly frequent that the stat section is just a statement of data
whereas appropriate significance permanet P less than 0.05 in that case
you have no idea you know and and what what is as appropriate especially when
your stats education isn't standardized that can be very different mean very
different thanks to very different people so one of the things that we're
pushing towards now is really encouraging investigators to report so I
can tell what stats were used I can tell I have the information there to verify
the test result and if there is a problem with investigators using a
statistical test that isn't relevant to their data I can pick up those problems
well okay so the question is whether there's any sense of how prevalent the
same problems are in social sciences with also have small sample sizes I
think there's a very active research group in psychology that's looking at
these issues and it's found many of the same types of variables and in fact for
things like bar graphs there are papers showing that practice is common in
psychology as well which we cited when we did our original paper so I've seen a
lot there I've seen less on the other social sciences but that may just be
because there isn't the same level of interest right now
psychology knows that they have a problem with through producing findings
and they're really committed to they have a core group of people that's
really committed to solving that problem so that's where I've seen the most
doesn't mean it's not happening elsewhere it's just based on what I read
that's what I see okay I talked to editors on a regular basis
one of the things so the question is any suggestions for editors about how to
approach your editorial board about getting changes made when I talk to
editors the ones that are able to make changes are often saying we have a lot
of people who like this we have some people who are dead set against it we
happen to have people in the right positions who like this and that's
allowing us to make the changes forward so I think that is a critical piece and
you know you know that's what the makeup of your board is I think presenting
information and papers that are really written for basic scientists are
important so ours is there's a couple of others that I can give you that are as
well I think emphasizing what other journals are doing so I have a list of
journals that have made policy changes with regards to their statistics that
have added checklists that are doing things on data visualization and so
checking other journal policies and how specifically their wording things and
what they're doing can also be something that encourages journal editorial board
members to make similar changes so those are all strategies that I would
recommend the other thing is if you can get people to make changes in their own
lab they're much more likely to be comfortable with encouraging others to
make those changes so we found that if we can get people to start graphing
their data in dot plots it fairly quickly becomes obvious to them why bar
graphs are such a problem and we don't have to do a lot of other explanation
explaining because they just get it so anything you can do to encourage journal
editors to try things out in their own lab before they push a policy forward I
think can also be helpful for convincing
from what I see in Physiology were so far away from having a problem and I'm
not too concerned about it that would be my answer I mean I think with the type
of literature that I deal with we are thinking about you know how can we get
people from to go from five to ten or 15 or 20 which is still within the
point where there's a lot of inaccuracy in those summary statistics where
there's low power unless you're looking for a really big difference we're not at
the point where we're you know thinking about 500 versus a thousand which is
where some of those other detecting really small differences that aren't
clinically meaningful become an issue yes it will definitely have an impact
and in its so we tend to think of power as being sample size power is not just
sample size it's also the amount of variability in your data so if you
reduce the variability then you increase your power and it's also the Mensch
between your groups so the bigger your difference between means Armenian is the
higher your power so when you're thinking about increasing power for a
small study it's not just increasing sample size and if you can't do that
there's nothing you can do there are other things and that may be one of
those areas where if you have large differences as well you do have
relatively high powered studies just because your variability is so low so I
would say Daniel Atkins so the question is about is there a statistics course
online that I would recommend Daniel Atkins has a good course through
Coursera and I don't remember what it's called offhand but I think it goes and
it's heared a little bit more towards a psychology audience but it is general
enough that it's still very useful for basic sciences so I think that has some
very good there's a lot of simulations and things to help you understand like P
curves how did those work how do they infect it affected by power and sample
size in those types of the start and then I think Twitter is
also a really good resource for getting information on all things meta research
so I would look there and start to find and follow meta researchers on Twitter
and they're going to be sending out papers and resources and links all the
time that can be very helpful to you
-------------------------------------------
Amin Younes Lifestyle , Net Worth, Salary, House, Cars , Awards, Education, Biography And Family - Duration: 3:17.
Please subscribe my channel
-------------------------------------------
Onslow Co. debuts asthma education kits to raise awareness about possible triggers - Duration: 0:48.
For more infomation >> Onslow Co. debuts asthma education kits to raise awareness about possible triggers - Duration: 0:48. -------------------------------------------
Zebrainy ABC Wonderlands - Learn N alphabet letters - Education Game App for Kid - Duration: 2:48.
Zebrainy ABC Wonderlands - Learn A to B alphabet letters - Education Game App for Kid
Zebrainy ABC Wonderlands - Learn A to B alphabet letters - Education Game App for Kid
Không có nhận xét nào:
Đăng nhận xét