Monetary unit sampling (or MUS) biases higher valued items. The likelihood of an item being selected is directly proportional to size of the item, unless that item is affected by top stratum cutoff. See Top Stratum Cutoff. With monetary unit sampling, a $1,000 item is 1,000 times more likely to be selected than a $1 item because a $1,000 item contains 1,000 times as many dollars, each of which is equally likely to be selected.
With monetary unit sampling, the data is treated as a stream of dollars, actually expressed in cents. Each dollar has an equal chance of selection and when a dollar is selected, the item containing the dollar is output. The population is the absolute value of all the dollars in the file. The interval you specify is the number of dollars between each selection.
This selection method inherently biases the sample towards greater dollar value items.
Use a monetary unit sampling base if you want to express a dollar confidence relating to the given population, for example, that total errors do not exceed $1,000,000. Monetary unit samples are useful for substantive or overstatement tests. By biasing higher value items, an MUS sample provides a high level of assurance that all significant items in the population are subject to testing. This is because when testing for overstatement, it is the higher value items that present the greatest risk of containing a material error.
The population for a monetary unit sample is the absolute value of the field being sampled. Fields can contain negative as well as positive numbers and negative numbers cannot result in overstatement.