Typo and/or missing info in Ch. 6 of MARK book

[trrobbin]

On page 6-23, after the 'end sidebar', at the end of the fourth paragraph down. We "change the number of parameters for this model from the 15 that are reported to the 16 that are actually estimated" Doing so then changes the results browser's calculation of QAICc etc.

The missing info, if I'm getting this right, is on page 6-30, the second paragraph after the figure. The fourth sentence reads "Using the Akaike weights, the model without the interaction is approximately 3.5 times as well supported . . . "

Do we not need to change the number of parameters for the reduced model as well? Doing this then changes the calculations in the results browser and the QAICc is closer to 2 times as well supported instead of 3.5 times. We may all get a little different number based on our bootstrapped c-hat adjustment, but I was thinking that the change in the number of parameters should be re-mentioned here, if I'm understanding this, and the calculations in the results browser may change a little. The LRT test results are still similar to what is in the MARK book, however.

Thank you for any clarification.

Travis

[cooch]

On page 6-23, after the 'end sidebar', at the end of the fourth paragraph down. We "change the number of parameters for this model from the 15 that are reported to the 16 that are actually estimated" Doing so then changes the results browser's calculation of QAICc etc.

The missing info, if I'm getting this right, is on page 6-30, the second paragraph after the figure. The fourth sentence reads "Using the Akaike weights, the model without the interaction is approximately 3.5 times as well supported . . . "

Do we not need to change the number of parameters for the reduced model as well? Doing this then changes the calculations in the results browser and the QAICc is closer to 2 times as well supported instead of 3.5 times. We may all get a little different number based on our bootstrapped c-hat adjustment, but I was thinking that the change in the number of parameters should be re-mentioned here, if I'm understanding this, and the calculations in the results browser may change a little. The LRT test results are still similar to what is in the MARK book, however.

Thank you for any clarification.

Travis

You are correct - for that particular example, all of the models end up needing some level of adjustment to the number of estimated parameters (some by more than 1, in fact - some of the constrained phi models actually result in >1 inestimable p). The work files I have at this end show that, but for some reason, that didn't make it into the text. I've just corrected Chapter 6, and uploaded a new version a few minutes ago.