From: Stan Hilliard
Date: 09 Apr 2000
Time: 00:47:36
Greetings Gregg,
Here are some suggestions concerning dollar unit sampling:
1) Make sure that the margin of error of the estimate of amount due is sufficiently small. Dollar unit sampling plans are attribute plans, which have a larger margin of error than variables plans.
2) Make sure that the sample selection method provides every dollar of the population an equal probability of being selected. A bias would be created if documents or line items were selected randomly without replacement.
3) Make sure that the sign of the error is taken into account. Not only chargeback, but also chargeforward should be possible.
4) Consider a strategy that requires an estimated chargeback or chargeforward to be significantly different from zero before implementation is allowed. You can use a hypothesis test to determine this.
Sincerely, Stan Hilliard