Occupancy groups and capture history

[TGrant]

I have two questions that I was wondering if anyone might have some insight:

For the occupancy analysis I'm working on now, we surveyed 5 meadows for a butterfly. Two meadows had no butterfly detections. But we had several plot covariates. I excluded the two meadows with no sightings for the analysis. Now I'm wondering if I should have included them because the plot covariates might inform the analysis.

On a future analysis, on another endangered butterfly, there is an issue with demographic closure. They have short flight season and the surveys extended through the end of the flight season, so detections dropped off to zero as the butterflies died off. I was going to model detection probability as a trend or quadratic trend to try to deal with this. Unfortunately the survey dates are all over the map and difficult to divide into discrete "occasions". I thought to try to just use each day as an occasion and just have lots of missing data between. Has anyone ever tried to run an analysis that had a large proportion of missing data? I have 707 surveys over a potential 8036 occasions. (164 plots by 49 days)

Thanks,
Tyler Grant
USFWS

[darryl]

You don't have to have the columns lining up chronologically for each plot (although p(t) models only make sense when you do), you could just have 1st survey, 2nd survey, etc. for each plot then keep track of which date a survey was conducted at each plot with a covariate(s). However, when you have a clear violation of the closure assumption (and it's not changing at random), there could be some bias creeping into your estimates, or it's not clear what the final estimates actually mean.
Darryl

[TGrant]

Thanks, that didn't occur to me to use date as a covariate.

I'll probably have to chop off the data at some reasonable point where there was still butterflies. I think there are 2 or 3 occasions where butterflies were still there.