Description

FOUR MODELS

. Unstratified Mean Per Unit

. Stratified Mean Per Unit

. Difference Estimation

. Ratio Estimation

CHARACTERISTICS (some, general):

. Answers the question how much

. Used to estimate an amount or test reasonableness of balance

. Quantifying (vs. qualifying under attribute models)

. Use on substantive tests

. Smaller sample sizes possible if many differences between recorded and audited amounts

. No "re-design" considerations if zero or negative items selected for "n"

. Easier to expand "n" if necessary as compared to PPS

. More complex than PPS

. Must be able to estimate Standard Deviation of "N"

. May not be appropriate if "n" is small and large differences exist between recorded and audited amounts

UNSTRATIFIED MEAN PER UNIT ESTIMATION (Simple Extension):

Calculate estimated amount for population (=estimated audit value) using sample mean

Examples: no book value available; footed total not accurate

Not as efficient as Stratified Mean per Unit if can stratify population or Difference or Ratio Estimation

Required to define materiality (precision) and reliability (confidence level)

Reliability/confidence level generally high (90 - 99%)

STRATIFIED MEAN PER UNIT ESTIMATION

Calculate estimated amount for population (=estimated audit value) using sample mean

Examples: no book value available; footed total not accurate

No or moderate monetary errors expected

Efficient (smaller sample sizes), but not as efficient as stratified Difference Estimation, Ratio Estimation or unstratified Difference Estimation

Define each stratum

Required to define materiality (precision) and reliability (confidence level)

Reliability/confidence level generally high (90 - 99%)

DIFFERENCE ESTIMATION

Calculate estimated audit value for "N"

Examples: book value available for each item; total book value known; differences expected

Differences between audited and book values must exist

Difference estimation may not be useful where strong internal controls exist

Differences should be both over and understatements

Over and understatement amounts should be relatively constant

"n" differences used to estimate total differences

Cannot use with low error rate populations (use discovery instead)

Most Efficient (smaller sample sizes) when stratified, along with ratio estimation;

Can stratify population or not Unstratified best when few differences (between audit and book value) Stratify (or use ratio estimation) when differences highly skewed Define each stratum Group similar items in same stratum (reduces variability)

Required to define materiality (precision) and reliability (confidence level)

Reliability/confidence level generally high (90 - 99%)

RATIO ESTIMATION

Similar to Difference Estimation

Calculate estimated audit value for N

Examples: book value available for each item; total book value known; differences expected

Differences between audited and book values must exist General rule: should have about 30 differences in final sample; if not use stratified MPU or PPS; cannot use if low error rate in N (use discovery instead)

Book values must be positive