I didn't follow all the sampling issues but part of the reason homogeneous
groups are used is because it reduces error variance. Like two group of white
rats who are so similar that even small effects become 'significant'. I.e. You
can't do both internal validity (does iv have ANY effect) and external validity
(who can we generalize the effects to).
The famous aspirin study (NE journal medicine) used 22000 men I believe. The
effect size was small (.9 vs 1.7% MI rates; though the relative risk is
impressive, almost 2x). But once the causal link is established other research
can follow up with other groups and/or diverse peoples.
I like omnicentric
Sent from my BlackBerry® wireless device from U.S. Cellular
------------------------------------------------------------------------------
I guess if we test for interaction effects using ANOVA and find a
significant ibteraction effect this blows out the assumption of minimal error
variance of homogeneity.
I agree that internal validity does not necessarily imply external validity.
Michael Sylvester,PhD
Daytona Beach,Florida
---
To make changes to your subscription contact:
Bill Southerly ([email protected])
