Questions about c-hat

[egmartins]

I have 3 questions concerning the use of c-hat. I look for the answers in "A gentle introduction" but I could not find them. Here they are:

1. C-hat used to adjust for lack of fit must be calculated only for the general model?

2. Should I always use c-hat to adjust for lack of fit, even when overall test 2+test 3 are non-significant, indicating that the ganeral model fits to the data?

3. What do I do when the estimate of c-hat based on the results from program RELEASE is less than 1 (in my analysis it was actually 0.75)? Should this value of c-hat be used to adjust for lack of fit or only values higher than 1 should be used?

Thanks very much in advance.

[cooch]

I have 3 questions concerning the use of c-hat. I look for the answers in "A gentle introduction" but I could not find them. Here they are:

1. C-hat used to adjust for lack of fit must be calculated only for the general model?

Yes, as it it says probably 3-4 times on the first page alone of the GOF chapter.

2. Should I always use c-hat to adjust for lack of fit, even when overall test 2+test 3 are non-significant, indicating that the ganeral model fits to the data?

It won't hurt - it is useful to compare/contrast results from RELEASE, bootstrap, median.

3. What do I do when the estimate of c-hat based on the results from program RELEASE is less than 1 (in my analysis it was actually 0.75)? Should this value of c-hat be used to adjust for lack of fit or only values higher than 1 should be used?

See top of page 3 - deals with this explicitly.


I'd suggest a more careful read of the documentation - including both 'the book', the helpfile, and other previously asked questions on this forum - before posting your questions.

[egmartins]

I have 3 questions concerning the use of c-hat. I look for the answers in "A gentle introduction" but I could not find them. Here they are:

1. C-hat used to adjust for lack of fit must be calculated only for the general model?

Yes, as it it says probably 3-4 times on the first page alone of the GOF chapter.

2. Should I always use c-hat to adjust for lack of fit, even when overall test 2+test 3 are non-significant, indicating that the ganeral model fits to the data?

It won't hurt - it is useful to compare/contrast results from RELEASE, bootstrap, median.

3. What do I do when the estimate of c-hat based on the results from program RELEASE is less than 1 (in my analysis it was actually 0.75)? Should this value of c-hat be used to adjust for lack of fit or only values higher than 1 should be used?

See top of page 3 - deals with this explicitly.


I'd suggest a more careful read of the documentation - including both 'the book', the helpfile, and other previously asked questions on this forum - before posting your questions.

Dear Cooch

Thanks very much for your message. I am sorry for the questions. I was not too certain about the answers to the first two questions when I read the book and I wrote just to confirm. Concerning the third question, I really read the top of page three but I did not remembered that at all when I finish the chapter. I do apologize for that.