advanced methods and analysis for the learning and social sciences
Download
Skip this Video
Download Presentation
Advanced Methods and Analysis for the Learning and Social Sciences

Loading in 2 Seconds...

play fullscreen
1 / 62

Advanced Methods and Analysis for the Learning and Social Sciences - PowerPoint PPT Presentation


  • 85 Views
  • Uploaded on

Advanced Methods and Analysis for the Learning and Social Sciences. PSY505 Spring term, 2012 April 26, 2012. Today’s Class. Frequentist Statistics not Frequently Covered in 21 st Century Textbooks. Why aren’t they covered?. Could cover a lot of methods. With a ton of math. Instead.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' Advanced Methods and Analysis for the Learning and Social Sciences' - emele


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
advanced methods and analysis for the learning and social sciences

Advanced Methods and Analysis for the Learning and Social Sciences

PSY505Spring term, 2012

April 26, 2012

today s class
Today’s Class
  • Frequentist Statistics not Frequently Covered in 21st Century Textbooks
instead
Instead
  • I’d like to try to make a different point
you don t need to know
You don’t need to know
  • The math behind every test
you don t need to know1
You don’t need to know
  • The math behind every test
  • Will you even remember it?
you don t need to know2
You don’t need to know
  • The math behind every test
  • Will you even remember it?
    • If so, you have a better memory than me!
what you need to know
What you need to know
  • What tests are out there
  • Enough math to be able to learn those tests when you need them
    • Many are in SSPS
    • Still good to know enough math to make sure that you are using them right
it s all out there
It’s all out there
  • By now, there is a test for almost everything you might want to do
  • Although it may be quite obscure
it s all out there1
It’s all out there
  • By now, there is a test for almost everything you might want to do
  • Although it may be quite obscure
  • The Ferguson book I excerpted from is a great resource
  • Another great resource is Sheskin’s Handbook of Parametric and Nonparametric Statistical Procedures
  • The Handbook of Research Synthesis by Cooper and Hedges is another nifty resource
the key thing
The key thing
  • Make sure you’re doing the right test
  • Right assumptions about
    • type of independent variables
    • type of dependent variables
    • independence or lack thereof
    • question you’re asking
if you don t know what you re doing
If you don’t know what you’re doing
  • Read some more
  • Ask somebody who has done similar tests in the past
so that said
So that said
  • Let’s play a game
so that said1
So that said
  • Let’s play a game
  • With candy!
put away your laptops
Put away your laptops
  • Or no candy for you!
what test do you use1
What test do you use?
  • If you want to compare means of two quantitative variables in two groups of students? (assuming no other variables as covariates, and that data is reasonably close to normal)
what test do you use2
What test do you use?
  • If you want to compare means of two quantitative variables in two groups of students? (assuming no other variables as covariates, and that data is reasonably close to normal)
  • Two group t-test
what test do you use3
What test do you use?
  • If you want to compare means of two quantitative variables in two groups of students? (assuming no other variables as covariates, and that data is not close to normal)
what test do you use4
What test do you use?
  • If you want to compare means of two quantitative variables in two groups of students? (assuming no other variables as covariates, and that data is not close to normal)
  • Wilcoxon Rank-Sum Test
  • Variant exists for two variables in one group
what test do you use5
What test do you use?
  • If you want to compare means of two quantitative variables in two groups of students? (assuming no other variables as covariates, and that data is not close to normal)
  • Wilcoxon Rank-Sum Test
  • Variant exists for two variables in one group
    • Wilcoxon Matched-Pairs Rank-Sum Test
what test do you use6
What test do you use?
  • If you want to compare means of two quantitative variables in the same group of students? (assuming no other variables as covariates)
what test do you use7
What test do you use?
  • If you want to compare means of two quantitative variables in the same group of students? (assuming no other variables as covariates)
  • Paired t-test
what test do you use8
What test do you use?
  • If you want to statistically determine whether inter-rater reliability is higher than chance?
what test do you use9
What test do you use?
  • If you want to statistically determine whether inter-rater reliability is higher than chance?
  • Friedman’s test(Sometimes called Friedman’s Chi-Squared)
what test do you use10
What test do you use?
  • If you want to statistically determine whether inter-rater reliability is higher than chance?
  • When might this case come up?
what test do you use11
What test do you use?
  • If you want to statistically determine whether one correlation is higher than another correlation? (For different populations)
what test do you use12
What test do you use?
  • If you want to statistically determine whether one correlation is higher than another correlation? (For different populations)
  • Significance of the Difference Between Two Correlation Coefficients for Independent Samples
what test do you use13
What test do you use?
  • If you want to statistically determine whether one correlation is higher than another correlation? (For different populations)
  • When might this case come up?
what test do you use14
What test do you use?
  • If you want to statistically determine whether one correlation is higher than another correlation? (For the same population)
what test do you use15
What test do you use?
  • If you want to statistically determine whether one correlation is higher than another correlation? (For the same population)
  • Significance of the Difference Between Two Correlation Coefficients for Correlated Samples
what test do you use16
What test do you use?
  • If you want to statistically determine whether one correlation is higher than another correlation? (For the same population)
  • When might this case come up?
what test do you use17
What test do you use?
  • If you want to compare two proportions in two groups of students? (assuming no other variables as covariates)
what test do you use18
What test do you use?
  • If you want to compare two proportions in two groups of students? (assuming no other variables as covariates)
  • Significance of the Difference Between Two Independent Proportions
  • OR
  • Chi-squared test, with df=1
what test do you use19
What test do you use?
  • If you want to compare two proportions in two groups of students? (assuming no other variables as covariates)
  • When might this case come up?
what test do you use20
What test do you use?
  • If you want to compare two proportions in the same group of students? (assuming no other variables as covariates)
what test do you use21
What test do you use?
  • If you want to compare two proportions in the same group of students? (assuming no other variables as covariates)
  • McNemar’s test
  • Also called the
  • Significance of the Difference Between Two Correlated Proportions
  • NOT Chi-squared!!!
what test do you use22
What test do you use?
  • If you want to compare two proportions in the same group of students? (assuming no other variables as covariates)
  • When might this case come up?
what test do you use23
What test do you use?
  • If you want to determine whether one variance is significantly different than another variance?
what test do you use24
What test do you use?
  • If you want to determine whether one variance is significantly different than another variance?
  • Significance of the Difference Between Variances for {Independent, Correlated} Samples
what test do you use25
What test do you use?
  • If you want to determine whether one variance is significantly different than another variance?
  • When might this case come up?
what test do you use26
What test do you use?
  • If you want to determine whether a process that would occur 50% of the time by chance, actually occurs more or less than 50%?
what test do you use27
What test do you use?
  • If you want to determine whether a process that would occur 50% of the time by chance, actually occurs more or less than 50%?
  • Sign test
what test do you use28
What test do you use?
  • If you want to determine whether a process that would occur 50% of the time by chance, actually occurs more or less than 50%?
  • When might this case come up?
what test do you use29
What test do you use?
  • If you want to determine whether a process that would occur 50% of the time by chance, actually occurs more or less than 50%?
  • http://www.youtube.com/watch?v=NbInZ5oJ0bc
other tests
Other Tests
  • Are there any situations you’ve come up with during your studies
  • Where there is a weird data situation
  • And you need an appropriate test
  • But you don’t know what to do?
  • Maybe I have some ideas…
    • Or one of your classmates
more than 10 m inutes left
More Than 10 Minutes Left?
  • Strube’s Adjusted Z!
last class
Last Class
  • Monday, April 30
  • 3pm-5pm
  • AK232
  • Ensemble Selection Methods
  • Special Guest Lecture: Zachary Pardos
  • Readings
  • Caruana, R., Niculescu-Mizil, A., Crew, G., Ksikes, A. (2004) Ensemble selection from libraries of models. Proceedings of the twenty first International Conference on Machine Learning.
  • Pardos, Z.A., Baker, R.S.J.d., Gowda, S.M., Heffernan, N.T. (in press) The Sum is Greater than the Parts: Ensembling Models of Student Knowledge in Educational Software. To appear in SIGKDD Explorations.
  • Assignments Due: None
final notes
Final Notes
  • This is the last class session I’ll be lecturing this semester
    • It has been a pleasure having each of you in this class and learning together with you
    • Thanks so much for being a great class
final notes1
Final Notes
  • Remember: When there’s a research question, there’s a good analytical or statistical method for it
final notes2
Final Notes
  • Remember: When there’s a research question, there’s a good analytical or statistical method for it
  • Occasionally you have to invent it
final notes3
Final Notes
  • Remember: When there’s a research question, there’s a good analytical or statistical method for it
  • Occasionally you have to invent it
  • But make sure to check first for a method that someone (or some research community) has already developed and refined
to put it another way1
To put it another way
  • “Anything that’s worth doing is worth doing badly” – Herb Simon
to put it another way2
To put it another way
  • “Anything that’s worth doing is worth doing badly” – Herb Simon
  • But gosh, it’s even better to do it well
    • And to be able to assess which way is better
final notes4
Final Notes
  • And if there’s one big idea from this class
  • Well, that’s it.
final notes5
Final Notes
  • This is likely to be my last class lecture as a WPI professor
    • Thanks so much for making this class a great capstone to my 3 years at WPI
final notes6
Final Notes
  • As you go forward, remember the most important Advanced Research Method
  • Do what you love
  • Do what you’re good at
  • And work really really hard
final notes7
Final Notes
  • http://www.youtube.com/watch?v=_hZrXdJ-ibo
ad