This macro constructs multivariable logistic regression models to enable you to test associations of the explanatory variables with the outcome after accounting for confounding and effect modification. Besides other statistics, it also presents you with the likelihood-ratio chi-square test statistics, Akaike's Information Criteria (AIC) and Schwartz Criteria (BIC or SC) for every model, thus assisting you in making informed decisions.
You may download this macro (sas) and theimplementation code (sas) but we strongly recommend you to work through the module and the worked example before implementing it. If you are new to model building, please also read a brief introduction to model building or consult other resources before you begin.
What it can do for you
The MultiLogistic macro constructs models using a forward, backward or stepwise approach (according to your specifications) and presents results in an Excel sheet (one row per model) to facilitate comparisons between models.
The output includes:
- Likelihood-ratio chi-square test statistics and P-values;
- Wald chi-square test statistics and P-values;
- Akaike's Information Criteria (AIC) and Schwartz Criteria (BIC or SC);
- Goodness-of-fit tests;
- Parameter estimates and standard errors of the study/exposure variable.
In addition, the macro can also construct and test interaction terms and create a table of final model results.
Work through the elearning module to learn how to implement this macro.