censoring data for a CJS model

[dd.snook]

Hello,

I'm analyzing encounter history data from a study on a migratory fish where radio-tagged fish were detected over 5 fixed receiver stations placed along the river. The study was repeated over 3 years but in one of the years the last two receiver stations were not operational. So I assumed that fish detected at the third station would have to be censored as shown in the encounter history below (the last three digits represent the year group):

10100 -1 0 0;

However I'm not sure how to deal with fish that was last detected for instance at the second station in that same year (encounter history: 11000). Should I censor them as well even though they could have been potentially detected at the third station? Or should I write the frequency of the encounter history as I would normally do without censoring? My concern in doing the latter is whether MARK will consider the last two stations when "writing" probability statements for that encounter history when it shouldn't as the fish could not be monitored downstream of the third station. Hope this is not confusing...

Thanks!

Ed

[Eric Janney]

Ed,

Can you provide some additional information about your study design? I'm guessing that you are using the CJS model to estimate "survival" over space instead of time periods to determine how the spawning run partitions out over a length of river? How many fish did you radio tag? I'm guessing that detection probabilities are very high (i.e., close to 1.0) when the transceivers were working properly? Are you pooling all of your data from the three years into one analysis and treating year as a group covariate? Your approach of censoring might work, or you could fix the last two p's to zero for that year group. At any rate, it will be difficult to compare the spawning run that year to the other two years because you won't be able to estimate "survival" for the reaches of river between the second and third array, or between the third and forth array.

Eric

[dd.snook]

Hello Eric,

Thanks for your reply! Yes, I'm interested in estimating survival over space for my three year groups (overall we tagged ~500 fish). We have some ad hoc estimates of detection probabilities that suggest they are very high indeed but not 1.0.

So would fixing the last two p's of one year group to zero work similarly to censoring the data at the third station? Thanks again.

Ed

[Eric Janney]

Ed,

I think either way would probably work, unfortunately you be able to compare that year to the other two.

Good luck,

Eric