22 Running Statistics

This chapter describes routines for computing running statistics, also known as online statistics, of data. These routines are suitable for handling large datasets for which it may be inconvenient or impractical to store in memory all at once. The data can be processed in a single pass, one point at a time. Each time a data point is added to the accumulator, internal parameters are updated in order to compute the current mean, variance, standard deviation, skewness, and kurtosis. These statistics are exact, and are updated with numerically stable single-pass algorithms. The median and arbitrary quantiles are also available, however these calculations use algorithms which provide approximations, and grow more accurate as more data is added to the accumulator.

The functions described in this chapter are declared in the header file gsl_rstat.h.