individual and group covariates in Multi-state models

[thaljoha]

I have a few questions about adding covariates. There are different ways to have time dependent models and add covariates, and I seem to have gotten a little confused during all of my reading. Below is an example of my coding for the design matrix for a multi-state model. I have three states: Pregnant (p), not pregnant (n), and calf (c). I have 30 years of data with weight and age covariates for each year in the study. I have made survival and recapture constant across groups and time. Also, I set P to C, N to C, and C to C all equal to the same parameter in the PIM chart and then fixed that probability to zero.

I believe that the probability of transitioning between states will change over time (depending on covariates of that year). So, I think I need one column for each type of transition for each year. So 90 total columns with t for the S and P, one for the psi P to C (which I set to zero), and then 29 columns for each of the other transitions (one for each year). Also, I added two columns of covariates-- age and weight-- which means that the effect of weight or age on the transition probability is constant through time. However I would need to make those two columns go diagonally if I want that effect to vary over time, correct? So, I would have 60 columns coding for my two covariates (30 for weight and 30 for age). In this scenario, would I keep my intercepts for the psi's the same (one in each column-- so 90 columns), or would I have a single intercept column for all of the psi's and two "coding" columns to identify the transition (but I think this would just state that there is a group effect by not time, so I don't need this, correct)?

Finally, how would I add group covariates (ie. weather, yearly population density, etc.) in the input file and the DM? Would it just follow the same format as individual covariates?

Thanks for all of your help