Linear constraints

[adamd]

I am having a problem with linear constraints. I am constraining resight probability to be a function of the amount of time that was spent searching for color bands. I first calculated search effort in minutes and entered these values. The resulting model had a huge deviance (500 times greater than other models in the set). I then recalculated resight effort in hours, days, and standardized effort on a scale of 0 to 1. Each of the resulting models had different deviance estimates, although the last 3 differed by a value less than 1. Shouldn’t all of these alternative model have exactly the same deviance, since they are constrained by the same linear function? Each of the linear functions are parallel (ie. have identical slopes).

Thanks for any help you can provide.

Adam Duerr
Graduate Research Assistant
Vermont Cooperative Fish and Wildlife Research Unit