A new R-function to estimate parameters of Gaussian mixture models simultaneously applying to multiple length-frequency data with different mixing ratios

Abstract

To analyze length-frequency data obtained simultaneously from multiple sites, we developed a new R function for estimating parameters of Gaussian mixture models that can be applied concurrently to multiple datasets with varying mixing ratios. The new function, “GMM_EM”, estimates parameters (means, standard deviations, and mixing ratios) by maximizing the log-likelihood of the Gaussian mixture model using the Expectation-Maximization (EM) algorism. The number of Gaussian components must be specified prior to running the function. To evaluate the performance of the new function, two datasets were analyzed: artificially generated length-frequency data and empirical length-frequency data for Ruditapes philippinarum obtained from 20 sites during a field survey at Yokohama-Umi-no-Koen (Marine Park). Results were compared with those obtained using the existing function “normalmixEM”.