covariate prediction confusion

[jlaake]

Sure. Look at ?model.average.list. It has an example. You need to get a list of the same estimates and vcv matrices from each model and then pass to model.average.

--jeff

[ked]

Thanks Jeff! The compute.real method works great, at least for a single model. I'm getting estimates and confidence intervals that match my attempts to do this by hand, which is great.

I'm now trying to figure out how to calculate prediction intervals rather than confidence intervals. Any ideas?
Thanks again!

[jlaake]

Please explain further because the only thing I can is relevant for this is a confidence interval unless you are thinking of something that you haven't explained. Prediction intervals in regression are for future observations.
What is it that you are trying to do. I'll be in the field away from email for the next week. Maybe someone else can help you with your question.

-jeff

[ked]

Yes, that's exactly what I'm trying to do - give a predicted survival rate under future environmental conditions, with the associated prediction interval. I suspect the interval would be really wide, but wanted to give it a shot based on some feedback I received on a manuscript. It seems relatively straight-forward to calculate for a basic linear regression model, but I'm not sure the best way to go about it for mark-recapture models.

[jlaake]

It isn't quite the same as regression because it is fitting the survivals to the line represented by your covariates. To get the equivalent you would have to fit a time model without the covariates and then use var.components or var.components.reml with the environmental covariates for fixed effects with a random effect error. Then you get a process variance which you can use in a prediction interval.

--jeff