Question & Answer
Question
I would like to assess the importance of variables in a Binary or Multinomial Logistic Regression. How can I obtain a Likelihood Ratio test in Binary Logistic?
Answer
Likelihood ratio tests can be obtained easily in either of two ways, which are outlined below.
First, you can use Binary Logistic Regression to estimate your model, but change the Method to Backward: LR (if using SPSS command syntax, the subcommand would be /METHOD=BSTEP(LR) followed by a list of independent variables). In the table "Model if Term Removed", consider the results for Step 1. For each variable, this will test the hypothesis that the full model (i.e. the model with all variables included) is indistinguishable from the model with that variable removed. Thus, the variable with the smallest significance has the most impact on the model.
Second, if you are using SPSS 9 or above, you can estimate the model using Multinomial Logistic Regression (NOMREG) instead. The output will include a table labelled "Likelihood Ratio Tests", which gives the same information as above.
Related Information
Historical Number
42075
Was this topic helpful?
Document Information
Modified date:
16 April 2020
UID
swg21479688