|
|
Hi all,
A simplified version of my problems is as follows:
1) I have a dependent variable with five categories (a, b, c, d, e) which
I'll call "outcome." For the multinomial analyses I used category "a" as
the reference category.
2) I have two predictors of interest, x and y (let's say that both are
continuous).
3) In the results I get both unstandardized coefficients and ORs for both X
and Y for each comparison (a-b, a-c, a-d, and a-e).
4) Now the question: I have an additional hypothesis about X for a
comparison not included in the results: b-e. I guess that the information
provided by the results using category "a" as the reference group should be
sufficient to estimate the effect of X when "b" is the reference group and
"e" is the examined category.
Does anyone know where I can find information that will help me determine
the regression coefficients of an independent variable for a comparison not
examined in the analyses (along with the significance test)?
I imagine that changing the syntax and use "b" as the reference category
would be a possibility but I have more than one hypotheses that involves a
couple of other categories so I don't want to spend a lot of time running
all possible combinations. In addition, I am concerned about potential
increase in type I error if I do all comparisons.
Any help will be greatly appreciated.
Thanks,
Pablo
|
Locked
