Clarification needed in Ch 5 (goodness of fit)?

[dhewitt]

Paul's message about Ch 13 reminded me about an issue in Ch 5 that might need some attention as well.

In the last full paragraph on pg 5-36, you have a discussion about why different c-hat estimation methods might differ. You implicate the link function as a reason for the difference in the bootstrap and RELEASE approaches, ultimately pointing to the number of estimated parameters. But the contingency tables used in the RELEASE method (Tests 2 and 3) don't depend on a link function or parameter counting, do they?

Regardless of this, I think Ch 5 is impressive and I refer to it all the time.

[cooch]

Paul's message about Ch 13 reminded me about an issue in Ch 5 that might need some attention as well.

In the last full paragraph on pg 5-36, you have a discussion about why different c-hat estimation methods might differ. You implicate the link function as a reason for the difference in the bootstrap and RELEASE approaches, ultimately pointing to the number of estimated parameters. But the contingency tables used in the RELEASE method (Tests 2 and 3) don't depend on a link function or parameter counting, do they?

No, they don't. In fact, I had deleted the offending text (one sentence, anyway) from my working version, which for some reason hasn't been uploaded. I'll do so soon.