PIM and design matrix

[mercedeherrera]

Hello everyone:

I am a new user learning program Mark. Now I am working with the examples of chapter 6 of the markbook.
In section 6.10 (>1 classification variables, n-way ANOVA) I am trying to build constrained models using PIMs and design matrix but I have a problem with the example multi-group.INP.
The model to run is:

phi=sex+colony+time+sex.time+colony.time+sex.colony+sex.colony.time

To simplify, it is assumed that p is the same for both sexes and both colonies, and constant over time. The model has 17 parameters (16 values for phi and 1 for p).

When I run the model built with the PIM chart and the logit link function, the AIC and nº of parameters are the same as in the markbook.
But when I build the model with the design matrix, although it is identical as in the markbook, after running this model, the results are different. I obtain only 13 parameters and AIC and deviance are higher. When I look at the real estimates I see that phi2, phi4, phi6 and phi8 are erroneous (close to 1.00) and p values from parameter 18 to parameter 32 are 0.5 (although the result browser shows only 13 parameters!) with standard errors and confidence intervals equals to 0.00.

Can someone tell me where I am wrong?

Thank you in advance for your help

Mercedes
University of Córdoba. Spain

[cooch]

But when I build the model with the design matrix, although it is identical as in the markbook, after running this model, the results are different.

Check again. If you built your DM to exactly match the one shown on *the bottom* of p. 57 of chapter 6, you'll end up with 17 parameters. I'd bet a significant amount of money that either (i) you mistakenly are thinking about/using the DM on the top of p. 57 (which does not yet have the 3-way interaction term), or (ii) you have something else awry -- perhaps starting some structures in the wrong row. It is relatively easy to get things 'out of alignment'.

[mercedeherrera]

Thank you very much for your kind and quick response.

I am very interested in learning program Mark to carry out a mark-recapture study in the framework of the European Management Plan of the endangered species Anguilla anguilla.
I have repeated several times the model, ensuring that the DM structure is the same as the one shown on p. 57, but regrettably I cannot find my mistake.

I hope to get lucky.

Thanks again for your support.

Mercedes Herrera
University of Córdoba
Spain