Effective sample size

From Wikipedia, the free encyclopedia
Jump to navigation Jump to search

In statistics, effective sample size is a notion defined for a sample from a distribution when the observations in the sample are correlated or weighted.[1]

Correlated observations[edit]

Suppose a sample of several observations is drawn from a distribution with mean and standard deviation . Then the mean of this distribution is estimated by the mean of the sample:

In that case, the variance of is given by

However, if the observations in the sample are correlated, then is somewhat higher. For instance, if all observations in the sample are completely correlated (), then regardless of .

The effective sample size is the unique value (not necessarily an integer) such that

is a function of the correlation between observations in the sample. Suppose that all the correlations are the same and nonnegative, i.e. if , then . In that case, if , then . Similarly, if then . More generally,

The case where the correlations are not uniform is somewhat more complicated. Note that if the correlation is negative, the effective sample size may be larger than the actual sample size. Similarly, it is possible to construct correlation matrices that have an even when all correlations are positive. Intuitively, may be thought of as the information content of the observed data.

Weighted samples[edit]

If the data has been weighted, then several observations composing a sample have been pulled from the distribution with effectively 100% correlation with some previous sample. In this case, the effect is known as Kish's Effective Sample Size[2]

References[edit]

  1. ^ Tom Leinster (December 18, 2014). "Effective Sample Size" (html).
  2. ^ "Design Effects and Effective Sample Size" (html).

Further reading[edit]