I work for the Missouri Department of Revenue where I manage the computer
assisted audit program for sale and use tax compliance audits.. Part of my
responsibility is involved with sampling on large audits. I am currently
working with a national work group sponsored by the Multisate Tax Commission
to develop a course syllabus for sampling.
The sampling that is done in tax compliance audits is mainly variables
sampling and in the computer audits we use statistical methods that may
include statistical evaluation. The most common method used is stratified
random sampling.
Accounting populations are very skewed and traditionally have a very low
error rate. We attempt of overcome these factors by using minimum sample
sizes that are larger than those suggested by variables and/or attribute
formulas and by stratification. Currently we use between 4 and 10 strata
with a top stratum consisting of 100% detail items. Stratum boundaries are
set by guestimate using the CV value as a guideline. For any given stratum
a CV value below .5 is acceptable, a CV greater that 1 indicates that we
must adjust. Our goals is to have each stratum as homogenous as possible.
I am looking for information on stratification approaches. I am aware of
approaches based on sum of squares of both the frequency and the dollars.
Specifically I am looking for information, algorithms, computer routines,
etc. for these approaches or any other approaches to optimal stratification.
Any assistance you can provide would be greatly appreciated...Thank you....
***********************************************************************
Joe K Evans CGFM
Manager Computer Audit
Field Audit Bureau St Louis
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|