MIXFIT: a microcomputer-based routine for fitting heterogeneous probability distribution functions to data

Some engineering data represent a mixture of data from two distinct populations or result from two or more concomitant phenomena. Homogeneous probability density functions are not always suitable for representing such data. This article describes the development of a procedure for fitting two types of heterogeneous density functions to mixed population data. The benefits derived from using these heterogeneous functions are evaluated with conventional goodness-of-fit tests. MIXFIT was used to fit mixed Gaussian mixed functions distributions to maximum flow rates from four gauging stations in Illinois, shown to be best represented by mixture distributions. In addition to evaluating the goodness-of-fit statistics for each distribution, MIXFIT was also used to estimate the 10-yr and 100-yr return-period peak-annual flow events. -from Authors