Confidence interval and sample design

bira

New Member
#1
Will the confidence interval change if the sample design changes from SRS to Within Cluster Sampling (used SRS within)
And both samples have the same size and mean
 

hlsmith

Not a robit
#2
If you are trying to balance confounders, and the latter does a better job, the estimate would be less biased and if its standard deviation is also minimized, then yes. The CIs are based on standard deviation, so keep that in mind. We can assume removal of some bias, but as for Std it would likely come down if given repeated execution of study but many little things could come in play such as what the relationships with the outcome and confounders and is they synergistic, and that association was erroneously being ascribed to the variable of interest which i would imagine to be another bias issue.