Dear Allstat,
It is well known that it is not worthwhile adjusting for clustering if the variable in
question has a very low intra class correlation (ICC) with respect to the clusters.
Does this also apply in the case of a regression analysis where the response (Y) and
explanatory variable of interest (X1) both have low values for the ICC but a third
variable (X2) which is not of direct interest but is included in the analysis for
adjustment, has a high value for the ICC?
Best wishes
David
David Boniface
Health Behaviour Unit
Department of Epidemiology and Public Health
University College London
2-16 Torrington Place
London WC1E 6BT
tel +44 (0)20 7679 6628
FAX +44 (0)20 7813 2848
|