in another part, if we find a association in overall analyse but not in the different strata of the stratifed analyse
You need to be much more specific about what your data looks like, what kind of model you are looking into (what kind of regression, what kind of strata, what kind of covariates?) You might also consider posting your question on StackExchange, which is perhaps a better place for the nuts-and-bolts of statistical analysis, or better yet talk to an on-staff epidemiologist or statistician.
WE CONDUCTED A CASE-CONTROL STUDY TO TEST WAHTEVER A POLYMORPHISM WAS ASSOCIATED WITH CANCER. IN OVERALL ANALYSiS WE FOUND AN ASSOCIATION. IN STRATIFIED ANALYSIS ACCORDING TO SMOKING STATUS WE TESTED THIS ASSOCITAION IN SMOKER APART AND NON-SMOKER APART.
i need to know how can i find the P for interaction by calculating the likelihood ratio
thinks
kacem
Find -2log L for the overall model without interaction G2 (model 1)
Find -2logl for the stratified model G2 (smoker) and G2 (non smoker)
G2 (Interaction smoking status with X)= G2 (model 1)- (G2 (smoker)+G2 (non smoker)), df=(2-1)*df of X. If X is continuous, df of X=1, if X is categorical with k level, df of X=k-1.
Doing above, the regression should only have smoking and X as predictors. If you have other factors, the above calculation can't be used. The stratified model will estimate the effect for other factors separately. The G2 difference will contain interactions with X and smoking*all other factors.
One more thing to say:
The G2 used in SAS is the -2logl for covariate only= the difference with and without covariates
Thinks Professor Yuanzhang Li for your helpful response. i find the
-2log L for the overall model without interaction G2 (model 1)
and -2logl for the stratified model G2 (smoker) and G2 (non smoker) using regression
but i dont understand what means G2 and in which model i find k term
and the calcul of df=(2-1)*df of X. If X is continuous, df of X=1, if X is categorical with k level, df of X=k-1
BEST REGARDS
KACEM
The df for the interaction of two categorical factors: A (with I levels) and B (with K levels) is (I-1)*(k-1).
the smoking status has two levels, hence I=2.
For two continuous factor, the df=1.
- Badges
- Science topic
- Similar topics
- Psychology