> Dear all
>
> I have three factors A, B and C, say, each of which has 3 levels: 1,
2 and 3, say. I wish to calculate the change in odds of having an effect
when an individual falls into level 1 of factors A and B and level 3 of
factor C as opposed to level 3 of factors A and B and level 1 of factor
C.
>
> I have therefore formed a two by two table of the form:
>
> No. with effect No. w/o effect
>
--------------------------------------------------------------------------------------------
> level 1 factor A,
> level 1 factor B n1 n2
> level 3 factor C
>
>
> level 3 factor A,
> level 3 factor B n3 n4
> level 1 factor C
>
-----------------------------------------------------------------------------------------------
>
> I have calculated the change in odds by hand according to the
formula:
>
> (n1/n2)/(n3/n4) [1]
>
> However, if I run a binary logistic regression analysis in SPSS in
which the only covariate included is factorA*factorB*factorC and have
checked that my reference categories are as wished, I obtain an
entirely different answer for the change in odds as defined by the Exp term.
>
> Why is this? I would be grateful for any necessary corrections.
>
> Many thanks in advance.
>
> Best wishes
>
> Margaret
---------------------------------
How much free photo storage do you get? Store your holiday snaps for FREE with Yahoo! Photos. Get Yahoo! Photos
|