Materiality threshold - Statistics newbie

Airek

New Member
Hi all,

I am not sure how to approach the following situation and am looking for advice on how to develop a quantitative measure -

The situation:
I am comparing a rate calculated from two different sources, but both have the same underlying data - numerator of this rate is a sub-population of the denominator population. The population (both numerator and denominator) should be equal in size, but may be different due to how the population data was gathered. The numerator being different receives MUCH more scrutiny than the denominator - several people ask why that number is different.

Example:
Database A - numerator is 100 and denominator is 2000, rate is 5.0%
Database B - numerator is 95 and denominator 2050, rate is 4.6%

How much of a difference can exist between the populations before the difference in the rate creates too much of an issue where investigation should be required to understand the difference? Should we focus on developing a threshold of difference for both the numerator and denominator or focus on the rate -- combination of both? How do I calculate relevant thresholds - forgive me, I am very new to statistical approaches so will likely ask several questions.