by simone77 » Thu Jun 10, 2021 3:58 am
You can set the interaction between two continuous covariates by adding a third column that is the product of the other two (numeric) columns. I don't see any problem with that, it is just what happens behind the scenes when you run a GLM-like model in whatever other program. To avoid large numbers that may be cumbersome for the computation, you may want to shift your values to get 0 equal to the minimum value or some other transformation that makes their product values smaller. An alternative is to multiply the covariates by a scaling value, this is suggested in CH 11.2.2 when they speak about individual covariate, I think it should work for environmental covariates too.
Of course, you must keep in mind what transformations you did in order to correctly interpret the results (i.e. the beta values). In the MARK book, there are several points about the interpretation of betas (I recommend that you also read chapter 11 - one may get MARK to treat environmental covariates as individual covariates indeed) and how this depends on the way the design matrix is constructed and the values of the covariates.