Multinomial Logistic Regression Analyses
The UniLogistic macro can also conduct multinomial logistic regression analyses when the outcome has more two categories that do not have any particular order. The procedure for conducting analyses is very similar to the binomial and ordinal logistic regression analyses discussed on previous pages, except the following statement:
modeltype = multinomial, /* Type of outcome variable */
Include this statement for conducting multinomial logistic regression analyses. Rest of the procedure is the same as for conducting binomial logistic regression analyses.
Example
The dataset MEEXP (Mammography Experience) is from a study conducted to assess factors associated with women’s attitudes towards mammography. The dataset was employed to demonstrate multinomial logistic regression analyses in Applied Logistic Regression by Hosmer and Lemeshow (2000). A brief description of the dataset ©John Wiley & Sons Inc is available at http://www.umass.edu/statdata/statdata/data/meexp.txt.
We have included a video showing the procedure to conduct multinomial logistic regression analyses for this dataset. You can download the dataset and the implementation code for conducting example analyses shown in the video. We have also included the output file produced by the macro.
- Download Video (wmv) (32.7MB; 2.09min)
- Dataset.
- Implementation code (sas).
- UniLogistic macro (sas) (You do not need to download it again if you already downloaded it for conducting binomial logistic regression analyses)
- Example output file (xls)
- Graphics - (Histograms (pdf), Box-plots (pdf), Bar charts (pdf), Stacked bar charts (pdf)).