limit on CLogit text?

[mtreglia]

Dear All,

In trying to fit resight probability to a linear trend I am in the process of entering in the parameter-specific link function values. I have two groups I am interested in (males and females), and 26 resight surveys, so I am trying to enter the CLogit function as:
CLogit(1,x)
CLogit(2,x)
Where x is the resight survey number. However, I cannot fit 2-digits in for "x" and still close the parentheses. Does it sound like I have my overall set-up incorrect? Or is MARK unable to fit that much text into these boxes?

Thank you!
Best,
Mike

[mtreglia]

Sorry to re-post with a follow up right away, but could I simply export the resight probabilities to a stats package and run an ANCOVA to account for how temp (the continuous measurement I'm interested in) and sex to affect resight probability? Or am I drastically missing something that MARK will tell me?
Thank you again!
Best,
Mike

[cooch]

Dear All,

In trying to fit resight probability to a linear trend I am in the process of entering in the parameter-specific link function values. I have two groups I am interested in (males and females), and 26 resight surveys, so I am trying to enter the CLogit function as:
CLogit(1,x)
CLogit(2,x)
Where x is the resight survey number. However, I cannot fit 2-digits in for "x" and still close the parentheses. Does it sound like I have my overall set-up incorrect? Or is MARK unable to fit that much text into these boxes?

Thank you!
Best,
Mike

Its a bug in the parameter-specific GUI that won't let you physically enter anything wider than CLogit(x,yy

Meaning, the right-most parentheses can't be entered (and, even if you try, isn't entered).

Fortunately, as long as yy<100 (i.e., is no more than 2 digits), this doesn't matter. If you enter

CLogit(1,10
CLogit(1,11

and so forth, in the .fpt file, this will show up as

CLogit(1,10;
CLogit(1,11;

and so on, instead of

CLogit(1,10);
CLogit(1,11);

But, it seems as though MARK parses the parameter-specific specification in a way that doesn't require the closing right-hand parenthesis. I suspect it simply starts parsing with the left hand parenthesis, reads the first field, hits the comma delimiter, then stops when it hits the semi-colon. At least, this seems to be the case, since it doesn't seem to affect the model deviance at all

CLogit(1,1);
CLogit(1,2);
CLogit(1,3);

gives the same deviances for a sample data set as

CLogit(1,10;
CLogit(1,11;
CLogit(1,12;


A simple bug Gary will no doubt fix at some point.

Now, the larger issue is whether or not CLogit makes sense. You mention trend. There are a number of ways to code for trend in MARK - some of them described in Chapter 6, some not.

[cooch]

Sorry to re-post with a follow up right away, but could I simply export the resight probabilities to a stats package and run an ANCOVA to account for how temp (the continuous measurement I'm interested in) and sex to affect resight probability? Or am I drastically missing something that MARK will tell me?
Thank you again!
Best,
Mike

No - because such a regression would not properly account for sampling covariances.

Back to your original post, if you interest is in the effect of temepreature, then all you need to do is include a model(s) where the parameter of interest is constrained to be a function of temperature. This sort of thing is discussed at some length in Chapter 6.

[mtreglia]

Thank you- I got hung up in looking at the estimated resight probabilities that are given in the output and forgot about the beta values. Also, your MARK book has been an invaluable resource!

Best,
Mike

[cooch]

Its a bug in the parameter-specific GUI that won't let you physically enter anything wider than CLogit(x,yy

Meaning, the right-most parentheses can't be entered (and, even if you try, isn't entered).

<snip>

A simple bug Gary will no doubt fix at some point.
.

You might be tempted as a short-cut solution to this (and similar) issues to directly edit the .fpt file (where the command code for a given model is stored). This is a binary file - although most of it is 'viewable' if you open the file in a simple ASCII editor, you can't use such an editor to 'fix' errors in the .fpt file. You can edit and save changes (say, adding the missing right parenthesis for the CLogit link commands) using said ASCII editor, but MARK will choke on the modified .fpt file.

You *can* successfully edit the .fpt file using a HEX/binary editor (Hex Editor Neo is a very good free HEX/binary editor), but you do need to know what you're doing (its actually pretty easy, but not something most people have much experience with).

[cooch]

Its a bug in the parameter-specific GUI that won't let you physically enter anything wider than CLogit(x,yy

Gary has just posted a new build of MARK which fixes this bug.

[mtreglia]

Thanks again Evan!
I'll download the new MARK build when I get a chance and look forward to playing with the code, now that you've given me enough information to dangerously spend hours trying to get things to work through programming as I try to learn the language
Cheers,
Mike