I first ask:
To compare two regression lines produced from two data sets, from spitting a
Data set into two categories, you just need to test B2 in the regression
Y=B0 + B1 (X) + B2 (D)
D is a categorical variable with dummy coding applied to the whole Data
(merging both data sets)
Hence, there is no need to test the interaction unless you want to lower the
value of "alpha" for the test
Is this true ?????????????????????
Most replies indecate that we should test the slop without restrecting the assumption of parral lines (comparing the mean for two groups).
However, we are still restricted to the normality assumption.
how can we test two generalized linear models without having to assume the same distribution for both regression lines?
Dr Ali Aljumaah
King Saud University
Riyadh, Saudi Arabia