Error: have you changed the data, because the saturated...

[simone77]

...model's likelihood has changed!

That's the warning message that pops up just before I can append the result to the results browser.

I am just playing around with my data and I am interested more in understanding how MARK works than in getting real estimates.
My dataset consist of more that 12000 individuals captured and resighted over 15 occasions. There are 6 groups. All the individuals are marked when they are young.
The model I have just tried is a TSM like this:

Phi(group*age2-t/t) p(group*age2-t/t)

There are 14 phi parameters for the juveniles of each group and 13 for the adults of each group. The same for p.
I have fixed 24 p to zero. In fact in two occasions there was no chance to capture or resight individuals (2* 2ages * 6groups= 24).

It is an overparameterized model but, as said, it is just to help me learning how the software works. According to the PIMs there are 324 parameters. The saturated model represents the perfect fit with one parameter for each data point.
So, this message makes ask myself: what does represent a data point in the MARK context? what does it mean that warning window?

Thanks in advance for any response

[gwhite]

Simone:
The reason you are getting this message is that you have changed the data type to a different one that has a different value for the saturated model's likelihood. This in itself is not a problem, but you have to be aware that the AICc values for the new data type are probably not compatible with the old data type. Hence, you cannot use AICc to evaluate which data type is the best, but only rank models within each data type.
Gary

[simone77]

Thank you Gary for your answer.
Anyway, I must be honest: I have no idea at what moment I have changed my data type. At least, I haven't done it by purpose. I believed I had just run a TSM model after running a CJS.
Anyway, I started a new session and (apparently) followed the same procedure and I got no warning message.