(Paul Anderson, New Scientist, 25 Sept 1969)
PsychStats is an informal and multidisciplinary group of psychology researchers (broadly interpreted) who are interested in statistics. Areas of research represented include education, emotion, intelligence, linguistics, personality, reasoning. Members have experience with a range of methods requiring overlapping knowledge (e.g., ANOVA, multiple regression, multilevel/mixed effects modelling, factor analysis, structural equation modelling), using a range of different software packages (e.g., SPSS, R, MPlus, SAS, Mx).
- have discussions drawing on research from the applied statistics literature;
- help each other to learn statistical methods and use them appropriately; and
- merrily ignore discipline boundaries.
We have a mailing list on which we discuss our troubles; this is also how the PsychStatsBanter gatherings (see below) self-organise and are advertised. If you would like to be added to the list, please mail Andy Fugard (a.fugard AT ed.ac.uk).
We meet on an occasional basis, usually in S38, Psychology, 7 George Square, with meetings lasting around an hour. Each meeting is focussed on issues in statistics that people find particularly troubling, socially and morally. To get an idea, check out past and future meetings below...
|Mon 23nd June|
|To be announced|
John Raven will come to chat with us about Progressing a Paradigm Shift in Psychometrics.
|Mon 2nd June|
|To be announced|
A discussion of Friedman, N. P., Miyake, A., Young, S. E., DeFries, J. C., Corley, R. P. & Hewitt, J. K. (2008) Individual Differences in Executive Functions Are Almost Entirely Genetic in Origin. Journal of Experimental Psychology: General, 137, 201-225. (Email Andy for a copy if you can't find it.)
|Thurs 24th April|
Anne will chat with us about SEM.
"I will provide a brief overview of SEM and describe some of my attempts to model the relationship between emotion and alerting (a type of attention) using AMOS. Issues relating to model specification, model identification, model fit and comparison will also be discussed. It is hoped that the overview and examples will be used a basis for discussion.
"All welcome - the brief overview might be of use to SEM newbies; additionally, it would be great if more experienced SEM-ers could come along to offer advice and clarification on murky issues!"
Ended up being a discussion of "multiplicative" effects vs. interactions (see Blanton and Jaccard, 2006)
|Thurs 6th March|
Discussion of Borsboom, D., Mellenbergh, G.J., & Van Heerden (2004) [The concept of validity. Psychological Review, 111, 1061-1071]
|Thurs 6th Dec|
|Tues 20th Nov|
|S32 (if numbers aren't too big!)|
A general discussion about Structural Equation Models.
|Weds 19th Sept|
Mike Allerhand will talk about likelihood: what it is, how to estimate maximum likelihood, using it for parameter estimation, and likelihood ratios for testing.
|Weds 22nd August|
We'll discuss model selection, e.g. AIC, BIC, likelihood ratio tests, stepwise methods, controversy, etc.
|Weds 8th August|
The structure of eyetracking data nicely highlights the deep angst faced by folk all over psychology so we're bound to acquire Deep Insight as a result of the struggle.
It'll help if you've read Baayen, Davidson and Bates (submitted).
|Weds 18 July|
Discussion of Gelman (2007) [Letter to the editors regarding some papers of Dr. Satoshi Kanazawa. Journal of Theoretical Biology, 245(3), 597-599] and the papers by Kanazawa it cites.
This covers lots of ground: interpreting logistic regression models, problems with multiple comparisons, properties of predictors to look out for in multiple regression.
|Fri 22 June|
Discussing Alex Weiss's and Alison Lenton's papers
|Wed 6 June|
|Alex Weiss's office (B18)|
... and which we may discuss soon! Email Andy (a.fugard AT ed.ac.uk) if you have more.
- More on mixed effects models... e.g. reporting models, analogues of Tukey's HSD, etc
- Simplifying terms in models, e.g. by merging levels
- Assumptions made when using covariates to "control" for something
- Signal detection theory
- Relationship between SEM, e.g. using them for latent growth models, and multilevel models
- Loglinear models and other models for categorical data
- Survival Analysis