Hi there,

I am doing known-fate models in Mark to estimate weekly winter survival of ruffed grouse. I'm using time-varying individual covariates with a logit link in my models. Some of my individual covariates are time-varying proportions and as such, range from 0 to 1. Others are time-varying environmental variables (e.g. snow depth and temperature) that have different value ranges that vary from week to week. I do not have any missing values.

One concept I've been struggling with is whether or not I should standardize the individual covariates. I've read pertinent sections about this in chapter 11 (Individual Covariates) and relevant pages in the help files (principally: https://sites.warnercnr.colostate.edu/g ... tes-basic/). First, I understand that Mark does some behind the scenes scaling of the design matrix (scaling each column so values range from 1- to 1) to ensure numerical convergence for estimates and that this can be turned off, if needed. I also understand that there is a box that you can check to standardize (z-transform) values across all individuals, but that there is caution around this when you have time-varying covariates and common intercepts-- I have both in my models. Because of this, I have NOT done any standardizing of my covariates either by checking the box in Mark or directly in my encounter history file before inputting into Mark. Throughout my reading however, there are sentences that start like "If you must standardize covariates..." and I still don't know exactly when someone might HAVE to standardize them.

All I can think of is if values are wide-ranging, but my understanding is that the internal scaling from Mark handles this. Or potentially, would I need to standardize if I wanted to create models with interactions between two continuous, time-varying covariates on different scales/units? If so, how would I safely do this? The help files suggests, "if you need to standardize the covariates, you must do so before the values are included in a MARK encounter histories input file, and you must use a common mean and standard deviation across the entire set of variables and observations." Would there be an issue with doing this if I have values like weekly snow depth that are 0 some weeks for all individuals and 50cm other weeks? Thanks so much in advance for the advice. -- Penelope