Occupancy robust design with two sets of points

[Nachoman]

Hello,

I am planning to use occupancy robust design models in RMark to estimate colonisation and extinction of patches within an experiment, but have some limitations with missing data, so I would appreciate some advice before starting the analysis. Briefly, the experimental design includes 4 experimental treatments with 24 replicate points per treatment, sampled every spring and autumn (with 5 visits in every sampling season) during 8 years. I plan to use treatment as grouping design covariate, and code year and season in the design matrix as covariates because they are of interest.

Now, I see two problems with the database that could affect estimation: i) the location of the points sampled for every treatment changed after 4 years for logistic reasons, so that we have data for one set of 24 points per treat for the first 3 years and another different set of 24 points per treat for the last 5 years; ii) we missed one season where no points at all were sampled in the 3rd year of sampling. So an ultra-summarized version of the dataset could look like:

10110 11011 11111 01001 ..... 11011 ..... ..... ..... ..... ..... ..... ..... ..... ..... .....
..... ..... ..... ..... ..... ..... 10101 11011 11111 01001 10110 11011 11111 01001 11111 01001

(line above first set of points, line below second set of points)


My question is whether RMark (or MARK) is able to deal with that kind of data, provided I am interested in treatment (and its interaction with covariates) effects, not in actual point estimates. As a less preferred alternative, I could analyse the first 3 years on one side, and the last 5 years indepedently (but still will have one missing occasion).

I would appreciate your views thanks. Also I hope this is the right forum to ask (a little bit confused since strong alternatives were MARK and PRESENCE forums!).

Thanks,

N

[egc]

Hello,

I am planning to use occupancy robust design models in RMark to estimate colonisation and extinction of patches within an experiment, but have some limitations with missing data, so I would appreciate some advice before starting the analysis. Briefly, the experimental design includes 4 experimental treatments with 24 replicate points per treatment, sampled every spring and autumn (with 5 visits in every sampling season) during 8 years. I plan to use treatment as grouping design covariate, and code year and season in the design matrix as covariates because they are of interest.

Now, I see two problems with the database that could affect estimation: i) the location of the points sampled for every treatment changed after 4 years for logistic reasons, so that we have data for one set of 24 points per treat for the first 3 years and another different set of 24 points per treat for the last 5 years; ii) we missed one season where no points at all were sampled in the 3rd year of sampling. So an ultra-summarized version of the dataset could look like:

10110 11011 11111 01001 ..... 11011 ..... ..... ..... ..... ..... ..... ..... ..... ..... .....
..... ..... ..... ..... ..... ..... 10101 11011 11111 01001 10110 11011 11111 01001 11111 01001

(line above first set of points, line below second set of points)


My question is whether RMark (or MARK) is able to deal with that kind of data, provided I am interested in treatment (and its interaction with covariates) effects, not in actual point estimates. As a less preferred alternative, I could analyse the first 3 years on one side, and the last 5 years indepedently (but still will have one missing occasion).

I would appreciate your views thanks. Also I hope this is the right forum to ask (a little bit confused since strong alternatives were MARK and PRESENCE forums!).

Thanks,

N

I'm going to move this to the general 'analysis & design questions' forum, since it really isn't 'software-specific'.

[Nachoman]

Thanks for that. Also, as an opportunistic aside, I would like to change my e-mail since my old e-mail address is no longer available (in case somebody wanted to discuss something outside the forum). I remember in a previous post that the administrator ask me to send a note, but I think I missed something since I can't find the email address where to send it (!) (contact details of site administrator "egc" are not available). Thanks, N.

[egc]

Thanks for that. Also, as an opportunistic aside, I would like to change my e-mail since my old e-mail address is no longer available (in case somebody wanted to discuss something outside the forum). I remember in a previous post that the administrator ask me to send a note, but I think I missed something since I can't find the email address where to send it (!) (contact details of site administrator "egc" are not available). Thanks, N.

Well, if I don't have your new email address, then its rather difficult to email you! Which is why you were sent the earlier email...which you apparently lost.

[egc]

Thanks for that. Also, as an opportunistic aside, I would like to change my e-mail since my old e-mail address is no longer available (in case somebody wanted to discuss something outside the forum). I remember in a previous post that the administrator ask me to send a note, but I think I missed something since I can't find the email address where to send it (!) (contact details of site administrator "egc" are not available). Thanks, N.

Moreover, without a functioning email address, replies to your posts will get lost, bounce back to the server, and cause the 'sysadmin' a few hassles. Best to avoid that.

Only viable solution is to delete your old account (saving your posts), and having you re-register.

Old account now deleted...re-register whenever you want.

[darryl]

Theoretically you should be able to analyse the data with RMark (or MARK or PRESENCE), your missing values just put some restrictions on the types of models you can fit. There may also be some confounding of temporal effects caused by your change in points (ie was a difference due to different time periods or different places) that you have to keep in mind when drawing your final inferences.
Cheers
Darryl

[nachoman]

Thanks for the advice, it has been very useful.Let's see how the analisys goes...

Cheers,

N