Polytomous (Multinomial) Logistic Regression

You will need at least 3 predictors of any type and one outcome variable that has several categories. Refer to the assigned readings, lecture materials, and workshop documents for guidance on interpreting the results of your analyses.

1. Provide a brief description of your data, including information about the sampling procedure and data collection method used.
2. Include a description of your variables. You should conduct univariate statistics and may wish to generate figures to complement information provided in table and text. As always, pay particular note to the distribution of your outcome variable.
3. Use the appropriate command to regress your categorical outcome variable on this one predictor. Using the examples provided in seminar, procure relative risk ratios. Interpret your output.
4. Change your base category to one that you feel is more intuitive. Interpret what you see.
5. Use the appropriate command to regress your categorical outcome variable on all of your predictors (Model 2). Interpret your output.
6. Drop one predictor from Model 2 (Model 3). Compare this third model with one fewer predictor (the ‘reduced’ model) to Model 2 (the ‘full’ model). Interpret the results of the likelihood-ratio test.
7. Based on the results of the likelihood-ratio test and other measures of model fit, select the best-fitting model. Provide evidence to support your choice, using examples provided in seminar.

Ordered Logistic Regression

You will need at least 3 predictors of any type and one outcome variable that has more than two ordered categories. Refer to the assigned readings, lecture materials, and workshop documents for guidance on interpreting the results of your analyses.

1. Include a description of your variables. You should conduct univariate statistics and may wish to generate figures to complement information provided in table and text. As always, pay particular note to the distribution of your outcome variable.
2. Use the appropriate command to regress your ordered outcome variable on this one predictor. Using the code provided in seminar, predict the relative risk and/or predicted probabilities for the outcome for any one of your predictors. Interpret your output.
3. Use the appropriate command to regress your ordered outcome variable on all of your predictors (Model 2). Interpret your output.
4. Drop one predictor from Model 2 (Model 3). Compare this third model with one fewer predictor (the ‘reduced’ model) to Model 2 (the ‘full’ model). Interpret the results of the likelihood-ratio test.
5. Based on the results of the likelihood-ratio test and other measures of model fit, select the best-fitting model. Provide evidence to support your choice, using examples provided in seminar.