MerQuR+Statistics Glossary

Statistics Glossary → Finite population correction (FPC)

Finite population correction (FPC)concept

A correction that shrinks the variance of estimates by the factor (1 − n/N) when, under sampling without replacement, the sample is a sizeable part of the population.

Standard error formulas usually assume an infinitely large population or sampling with replacement. In reality, when a sample is drawn without replacement from a finite population, uncertainty falls as the sample approaches the population size; if the whole population is counted (n = N), sampling error is zero. This is taken into account by multiplying the variance by (1 − f), where f = n/N is the sampling fraction. In stratified sampling the correction is applied within each stratum with its own fraction, and in multistage sampling at the level of the primary sampling units.

When the sampling fraction is small (roughly below 5%), the effect of the correction is negligible. The correction makes sense only for the uncertainty of descriptive estimates (the mean or total of this population); it is usually not applied when results are meant to generalize to a process beyond this population.

An everyday example

A company with 800 employees surveys 400 of them about job satisfaction. With a 50% sampling fraction, the variance is multiplied by 1 − 0.5 = 0.5 and the standard error shrinks by about 29% (√0.5 ≈ 0.71). Had the same 400 people been drawn from a city of a million, the correction would change practically nothing.

In MerQur

In the Weighted Mean, Weighted Frequency, Weighted Total and Weighted Linear and Logistic Regression analyses, the 'FPC (population)' column gives the stratum population size N_h and the variance is multiplied by (1 − n_h/N_h); in Survey-PHREG, the 'Finite population correction (FPC)' value can be entered as a sampling fraction or as the population number of PSUs.

Analyses where this concept appears

Weighted Mean (Survey Means)Weighted Frequency (Survey Freq)Weighted Total (Survey Total)Weighted Linear Regression (Survey Regression)Weighted Logistic Regression (Survey Logistic Regression)Survey-PHREG (Weighted Cox)

Related concepts

Stratified sampling (design stratum)