Applied Statistical Data Analysis in Psychology II
PSY 1952
Subject & Catalog Number
Course Information
Description
This PSY 1950 follow-up course continues with exposure to state-of-the-art Bayesian and Frequen- tist statistical modeling techniques. We begin by showing how the models we learned in Psy 1950 can be used for prediction. The next two units address power analysis (including simulations for mixed-effects setups), followed by two units on handling missing values. Subsequently, we introduce model comparison, including segmented and polynomial regression, and move on to a modern framework for modeling the effects of predictors in a truly nonlinear fashion:
generalized additive models (GAMs). GAMs can be seen as an extension of the GLMM framework. The same applies to regression shrinkage techniques like lasso and ridge, which incorporate penalties into regression, and can be used for predictor selection or in cases of multicollinearity.
Next, we explore path models, which enable structural regression extensions that go beyond standard "multiple IVs, single DV" settings. A popular example of such structural extensions is mediation models, which we cover in great detail, including longitudinal mediation.
This is followed by units on causal inference, which focus on making causal statements about the effect of one variable on another. We cover the potential outcomes framework and revisit mediation within this context. Finally, we take a deeper dive into Bayesian inference, including probability theory related to Bayes' theorem, ROPE testing, and MCMC convergence checks. We also learn how to define our own priors. The full topic list is provided below.
Recommended Prep
The Psychology Department requires completion of PSY 1950 or equivalent.
Available for Harvard Cross Registration
NOTE: This course requires additional sections; you will be prompted to choose secondary components during the Add to Cart process
