Significance tests


I'm working on basic data profiles for sample groups - volume, proportion percentage and index to universe - and am using z-scores to define the confidence in any significant results. However as the samples i'm working with for certain groups are limited i'm finding that indexes and z-scores aren't enough when the volumes are very small.

I need another stat to define when sample size is too small to consider robust so i can eliminate certain findings.

Any thoughts ?

Many thanks,