Incorporate a third species into 2 species models

[Bill]

Hello all,
I am hoping somebody can please help. I am running a two species model. I can only include the occupancy matrix of a third species as a survey covariate if I replace all of the 'not surveyed' times ie ('.') with '0'. Is this normal? I have modelled the third species as a seasonal covariate by condensing all of the individual surveys for the third species into a season by assigning a presence or absence score for the third species at each site for a season and then initiating a seasonal DM and including the associated seasonal covariate, however this model has very little support compared to the former. Is there a trick to including the '.' in the survey covariate matrix?
Regards, Bill.

[darryl]

Hi Bill,
You're post is missing a few details, like which parameters were you trying to include the 3rd-species covariates for. Also, you post appears different to what I received in my email inbox over the weekend so I presume you must have edited your question. As with any type of survey-specific covariate, if you have a 0 or 1 recorded for the detection data for the target species, there should a corresponding real value for your covariate, ie not a '.'. If you do have a '.' for the covariate PRESENCE should throw a warning. PRESENCE will play nicely if you have a '.' for the corresponding survey for both the covariate and the detection data. If you think that should be what you have, you might want to recheck your data.
Cheers
Darryl

[Bill]

Hi Darryl,
Sincere thanks for your reply. I am modelling a third species as a survey covariate for detection in a two species analysis using the first paramaterisation. My survey covariate data matches my presence/absence data, and Presence appears to play nicely unless it gets to a first season where the survey covariate starts with a '.' If the survey covariate starts with a '1' or '0' it appears to read the data, even when there are '.' following the first record for the first season. The program runs nicely when I replace all survey covariate '.' with a '0'. Given there is no associated presence/absence data for the '.' that I am replacing with a '0' in the survey covariate file, then would it be fair to assume that this manipulation does not affect the results?
Cheers,
Bill.

[darryl]

Hi Bill,
As a slight aside, firstly the first parametrization can be a bit unstable with covariates; the 2nd or 3rd parametrization are much better. If there is a "." for the corresponding detection survey it shouldn't matter what value you have in your corresponding survey-specific covariate as that value should be used. You could verify this being changing the value and see if you get the same result (which you should). Have you checked that you really did have "." in the covariate table and not and extra space like " ." or ". "? There's no reason why PRESENCE should get thrown by a "." in the first survey, unless there's a coding error for that particular model. Try fitting a single-season custom model and see if PRESENCE appears to run ok
Cheers
Darryl

[tajensen]

Have you tried replacing the " . " with a " - " ? I used the " - " to indicate missing values with no problems.

Tim