Separation of a mixture distribution into its Gaussian component

An application of an algorithm presented by J. Gregor in 1969 for decomposition of a mixture distribution into Gaussian components is presented here in relation to discrete mixture distributions of constant class interval. This graphical separation called Modified Gregor method (MG) uses the local modes to estimate the mean of components, the area around the mean for estimating the variance and subgroup population components. The Kolmogorov-Smirnov statistic is used for goodness of fit test of the composite distribution. This deterministic method has proven to be comparable to and as efficient as the commonly used Bhattacharya method. For comparison, the MG and Bhattacharya methodss are used in three examples related to fish population studies.