I am trying to run an ordinal logistic regression using Generalised Linear Model from SPSS. There variables are A,B,C,D, and a moderation variable A*C. Variables A and D are continuous variables while B and C are categorical variables. In the SPSS output, the "Test of model effects" table gives p-values of 0.005, 0.343, 0.372, 0.035, and 0.212 for variables A,B,C,D and A*C respectively. However, in the "Parameter Estimates" table, while the p-values remained unchanged for most of the variables, the p-value for variable A is 0.285 and its corresponding exponentiated estimate is 1.05 (95% CI=0.96, 1.14). Should I conclude that variable A as not statistically significant? May I know the cause of these contradictory p-values? Thanks.
I admit that has me a little bit stumped. However, I THINK that the 'Tests of Model Effects' tests the 0-order relationship between each IV and the DV, but that the 'parameter estimates' assess the full model. If so, that means that there is a 0-order effect between A and the DV, but that disappears when A is entered into the full model. That could be the case if A is correlated with the other IVs (probably D). Check the Pearson Correlation between A & D. Perhaps someone else can chime in.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation