Hi everyone,
I am a medical student and relatively new into statistics, I hope you can help me out with this question.
I'm using logistic regression to test whether there is an association between my main predictor (continuous) and the outcome (unadjusted model) and I'm also using an adjusted model to which I added several independent variables (IVs), such as continuous covariates (age and IQ).
The IV 'Age' in the adjusted model is not a significant predictor (p=0.681). However, when adding all interactionterms to test for linearity of the logit, "Age*LnAge" appears to be significant (p= .011).
I tested the linearity of the logit for all continuous IVs, using Andy Field's Discovering Statistics using IBM SPSS Statistics (4th ed.) As stated in 19.8.1. Testing for linearity of the logit, I transformed my continuous IVs to the natural log of the IVs (LnIV), and added the interactionterms of all IVs to the model (IV*LnIV) so check for any significant (< .05) interactionterms.
Is this a problem for my main predictor in the analysis? Or is this not relevant as age in the adjusted model is not a significant predictor on itself?
Thank you in advance!
Linearity of the logit has to do with only with the DV. This definition was completely confused by the SPSS literature. It has to do with psychological measures and in my opinion is not applicable to your situation..you might see the responses of Professor Kelvyn Jones to similar questions on RG. Best wishes, David Booth
See also. The discussion in Frank Harrel's Regression Modeling Strategies 2nd ed available in the z-library. Best wishes, David Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics