MULTIVARIATE LOGISTIC

The multivariate logistic model is, in its simplest form, a straight forward generalization of the univariate model.

Specifying multivariate models in PROC LOGISTIC is also straight forward

Homework
Use the data set lipid2018_b to estimate a multivariate logistic model that has chd as the dependent variable and any five of the other variables as independent variables.

The idea of confounding appears often in multivariate modeling.

Homework
Use the data set lipid2018_b to estimate the following models relating the specified variables to chd.
sbp as the only independent variable.
age as the only independent variable.
age and sbp as independent variables in the same model.
Restrict your displayed results to the parameter estimates.

The likelihood ratio test is included as a global test of the model in PROC LOGISTIC

The more general form of the likelihood ratio test allows testing a subset of the included parameters.

Proc genmod provides a "Type 1" analysis of likelihood ratio statistics.

Homework
Using the same five independent variables you selected above, use the data set lipid2018_b and proc genmod to conduct a "type 1" analysis of the likelihood ratio statistics.
Repeat the analysis, but change the order the variables in the model are specified and compare the results.

Confidence intervals are calculated in the same manner, but have a new interpretation.

Homework
Using the same five independent variables you selected above, use the data set lipid2018_b and re-estimate the model including 95% Wald based confidence intervals in the output.
Additionally, include a units statement that causes the procedure to calculate odds ratios for 1 unit and a standard deviation unit for all five variables.

The slides used in the videos are found here