I'm fitting a simple logistic model. Two covariates are described throughout a bivariate spline function defined with truncated linear basis (1 knots each). In this case I have:
Assuming that I want to check if the knot on one covariate is needed or not, since theoretically the two models are nested, the model above becomes:
Or should I also remove the last term (the interaction between the contrast and the other covariate)?
Thanks for your help.