How to assess variance for sample size calculation

#1
Hi everyone,
I am going to calculate sample size for a clinical trial. I need to compare two groups by quantitative variable. I have data of a previous phase trial: Ns, means and SDs for each group. What would be better way to assess variance for using formula for sample size: to take biggest value from two SDs or to calculate weighted average of them (Cohen's formula)? Or does anyone know the better approach?
Many thanks in advance!
 
#2
I would like to add that if mean difference between two groups is "big", then variance of combined sample should be bigger than variances of each groups, so both approaches are not appropriate.
 

hlsmith

Not a robit
#3
I didn't follow you r last part about how big mean difference would imply bigger pooled variance. Will exposure be randomized in trial so you will just be conducting a bivariate comparison?
 
#4
Hi, hlsmith,
thank you for your answer! I meant variance of combined sample. For example, efficacy variable in the 1st group has values 1, 2, 3 with mean=2 and SD=1, for the 2nd group: 11, 12, 13 with mean 12 and SD=1. Then SD for combined sample (1,2,3,11,12,13) is 5.55. Probably I don't understand something but I think that I should use 5.55 for calculation of sample size.