Question & Answer
Question
I ran a logistic regression analysis with the SPSS Logistic Regression procedure. The predictors included a categorical variable with 4 categories. The "Variables in the Equation" table in the output displays three coefficients for the 3 indicator parameters for this predictor. However, these are preceded by a row with the predictor name in the parameter name column. In this row, there is no B coefficient, S.E., or Exp(B), but there is a Wald value with 3 df and significance. What do the Wald value and significance for this row represent?
Answer
When a K-level predictor is specified as categorical in Logistic Regression, the procedure generates K-1 internal contrast variables. By default, these are indicators for the first K-1 categories. In the "Variables in the Equation" Table, there is a coefficient for each of these indicator variables, along with a Wald test and significance value. These are preceded by a test of the significance of the predictor as a whole. The p-value for the overall test will be based on a chi-square distribution with K-1 degrees of freedom. Suppose that you have a 4-level predictor called XCAT and that you specified it as categorical with indicator contrasts. Suppose also that you left the 4th category as the reference category. The first column in the "Variables in the Equation" table would represent XCAT as:
xcat
xcat(1)
xcat(2)
xcat(3)
There will be no parameter coefficient (i.e. no B, S.E., or Exp(B)) for the XCAT row. There will be a Wald test of XCAT's contribution to the model as a set of 3 indicator variables. The rows with XCAT(1), XCAT(2), and XCAT(3) will represent the 3 indicator variables and will have parameter coefficients that are part of the model.
Related Information
Historical Number
36536
Was this topic helpful?
Document Information
Modified date:
16 April 2020
UID
swg21475019