Instruments used to collect information on outcome and confounding variables need to be pretested, validated, and tested for reliability. One must ensure that any regularly collected data one uses (cancer registries, vitals statistics, birth outcomes from birth certificates) is accurate.

Statistical power is likely a very important issues as many investigations occur where populations are sparse.