I am in the process of creating an automated audit tool that does audit of System Information based on some set parameters. This involves three aspects - firstly identifying the exception in the populaton (for instance instead of a receipt, we have a payment), then calculating the sample size and then investigation of exceptions. The current sample size calculation based on Normal distribution (like most other sampling calculators) and it is based on the assumption of central limit theorom. However, the base data under question may not be a normal distribution. The challenge in that case is that we may be incorrect in our sample size calculation which may make our sample non-representative or make it much larger causing inefficiency(we are not staffed to deal with the substantially larger samples).

Is there a method through which I could adapt the calculation to the underlying distribution in an automated manner?

Hello Sridhar

 

What you are referring to is an automated tool that could help you identify the data distribution and then confirm the sample size. I am not sure that such approach would be useful. Every data set is unique and you will need a pair of eyes to look through and make sense of the data.

 

Regards

Mayank Gupta