Time-varying individual covariates

[harding]

Hello,

I am working with a colleague on a dataset involving a population of animals in which individuals are marked annually and re-detected weekly throughout the year (by radio telemetry or resighting). By the end of the study we will have 4 years of monitoring. During the annual capture period, which occurs just after the young of the year are born, all individuals are captured (new and previously marked animals; we have near 100% capture rates for this) and data including weight is recorded for each. A health status is also recorded throughout the year based on observations. We are interested in determining if survival varies with weight and/or health of individuals.

We are considering using a multistate analysis where the states reflect the health status. We also want to include weight as an individual time-varying covariate. I consider myself an intermediate-level MARK (through RMark almost exclusively) but have not previously dealt with individual covariates. After reading the guide and relevant archived posts, I am unclear about how to handle individual time-varying covariates for this sort of design. Am I correct in understanding that we would need a weight value for all 208 weekly sampling periods? Clearly we do not have such data. Would it be a problem to simply include weight as 4 separate individual covariates?

We have also considered breaking the data down by year and analyzing each year separately, starting with the period during which weights were recorded and using that weight as an a single individual covariate. Would this be a more appropriate way of handling the data given the study's design?

I understand that time-varying individual covariates are often best handled through a discretized MS framework, but if we were to further break down heath state into weight categories, I believe this would leave us with more states than we have data to run.... I suppose we could run MS models separately for weight categories and health status, but we are also interested in seeing if/how the two may interact.


Thanks in advance for any suggestions or advice!