Introduction to Categorical & Nonnormal Data Analysis

Assignments

Assignment #1 -- Using the ICU dataset, build and interpret a binary response model using status (sta) as the response variable. Check fit, assumptions and for influential observations.
Assignment #2 -- Use the lahigh dataset (see codebook) with math grade (math) as the response variable. Build and interpret an ordered logit model. Does the proportional odds assumption seem to hold? Using the same covariates, build a multinomial logit model. How do the coefficients, relative risk ratios and predicted probabilities compare in the two models?
Assignment #3 -- Use the death penalty dataset (deathpen.dta) with execution (execute) as the response variable. Build and interpret both a count model and a zero inflated count model. Which model fits better? How do the models differ?
Assignment #4 -- Rerun the model from assignment #3 using the generalized linear models approach. In glm it is possible to change the link function and/or the distribution family. Can you develop a "better" model by changing one or both of these options?