[md27]

Hi Everyone,

Could somebody please explain variable sampling, specifically MPU, as the logic doesn’t quite make sense to me. We are basically assuming there’s a misstatement if the mean dollar amount of our selected samples multiplied by the total items in the population is outside of our tolerable misstatement. I feel like this testing is awfully dependent on the samples we select. Does setting a threshold for tolerable misstatement really justify this type of testing? Also, what’s the point of making a projection of what the population should be when we have the population right in front of us when we make selections? Maybe I just don’t understand it yet, but I think if I was the client getting proposed to make an adjustment based off this rationale, I would tell them to screw off and test more.

[md27]

Also, what if we “randomly” decided to choose most of the low value items in the population as part of our samples. It would drive the mean down making it look like there's a huge misstatement when in reality there isn't. Like I said, this seems awfully dependent on the selections you make. I'm hoping somebody can help shed some light on this!

[Tim]

It's statistics. You will never be certain but you can state a conclusion with a 95% confidence level or whatever amount of confidence you think you need (professional judgment) to have “reasonable assurance” of detecting material misstatements. The reason they don't just sample the whole population is because that would be ridiculous in terms of cost/benefit when there's thousands or even hundreds of thousands of items in the population.

[md27]

Thank you for your response. I get what you're saying. I was just thinking that the selections you make are crucial for this testing as I could select a ton of the low value items in the population, test them, have no discrepancies in our selections with the book value and audited value, and still end up with a variance when projecting the total population. But I guess that's why we set a confidence level and tolerable misstatement….

[TCav12579]

The issue with your argument of selecting the smaller dollar balances is that you would not hit the necessary $ value to ensure you have tested enough to hit your performance materiality.

[md27]

Not sure I follow. Don't we have a set number of samples while having to ensure each selection is random? What if we “randomly” picked the low value items. If we're testing until the untested balance is below PM, that sounds like a high dollar value target testing, which means you can't use that test to project to the entire population.

[DoubleBogey]

That is why MPU is stratified, selecting all material amounts independent of the sample so they don't get overlook or not tested.

[md27]

Thanks @jsdailey. That helps clear up some things for me. I think what I was referring to earlier sounds a lot like sampling risk and the risk of incorrect rejection.

[Author]

Replies