Apologies if this is a thread which has been on allstat before. Are there any 'rules of thumb' for sample sizing Multiple regression and Analysis of covariance? I have an awful lot of unknowns for any precise estimate but prediction/confounding play a big part of a proposed study and I would like to say how many subjects would be appropriate. Results will be posted Thanks Paul ----------------------------------- Paul Wicks Senior statistician [log in to unmask] St. Georges Hospital Medical School %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%