Testing linearity in the logit using the Box-Tidwell transformation in SPSS (Part 2 of 2)

This video is part 2 of a 2-part series on how to perform the Box-Tidwell transformation and to use the transformed variables to test the assumption of linearity in the logit. Part 1 ( • Testing linearity in t... ) covered the basic concepts and strategy. This video picks up where the other vide left off by addressing a problem that could emerge when performing the transformation on predictors that contain values of 0 or negative numbers. A work-around strategy is provided in this video.

[@andreasullivan4412]

Thank you SO very much for these videos. When testing for linearity after the transformations of the continuous variables: if our model also has nominal variables, do we include all (continuous original, continuous transformed, and nominal) for the assumption significance test? or do we only need to include the continuous variables (original and transformations) in the test (despite other variables in our full model)?

[@InspirationIsFree]

Thanks for this video. If my independent variable has a minimum value of 0 and then only positive numbers, is this an issue? How many zero values would make it difficult to assess linearity?

[@mikecrowson2462]

Hi there. Yes, this could be a problem when testing for linearity in the logit because you cannot take the natural log of 0 (your minimum value). If you want to retain all your cases when testing for linearity in the logit, then you'd need to make sure the minimum value on your variable is > 0. Hope this helps!