- Pivotal quantity
In
statistics , a pivotal quantity is a function of observations whose distribution does not depend on unknownparameter s.More formally, given an
independent and identically distributed sample from a distribution with parameter , a function is a pivotal quantity if the distribution of is independent of .It is relatively easy to construct pivots for location and scale parameters: for the former we form differences, for the latter ratios.
Pivotal quantities provide one method of constructing
confidence interval s, and use of pivotal quantities improves performance of the bootstrap.Example 1
Given independent, identically distributed (i.i.d.) observations from the
normal distribution with unknown mean and variance , a pivotal quantity can be obtained from the function::where :and:are unbiased estimates of and , respectively. The function is the Student's t-statistic for a new value , to be drawn from the same population as the already observed set of values .Using the function becomes a pivotal quantity, which is also distributed by the
Student's t-distribution with degrees of freedom. As required, even though appears as an argument to the function , the distribution of does not depend on the parameters or of the normal probability distribution that governs the observations .Example 2
In more complicated cases, it is impossible to construct exact pivots. However, having approximate pivots improves convergence to
asymptotic normality .Suppose a sample of size of vectors is taken from bivariate
normal distribution with unknowncorrelation . An estimator of is the sample (Pearson, moment) correlation:where aresample variance s of and . Being aU-statistic , will have an asymptotically normal distribution::.However, avariance stabilizing transformation :known as Fisher's transformation of the correlation coefficient allows to make the distribution of asymptotically independent of unknown parameters::where is the corresponding population parameter. For finite samples sizes , the random variable will have distribution closer to normal than that of . Even closer approximation to normality will be achieved by using the exact variance:References
Shao, J (2003) "Mathematical Statistics", Springer, New York. ISBN 978-0-387-95382-3 (Section 7.1)
Wikimedia Foundation. 2010.