Mixed distributions -mixtures of normal distributions- arise frequently in biological and ecological data, and the analysis of these mixed distributions is an important topic in fisheries science, particularly when studying species in subtropical and tropical areas. Many samples of fish lengths for these species exhibit multiple modes. Since Petersen (1892), it has been recognized that these modes are evidence that the sample contains a mixture of several generations of fish. A variety of methods, both graphical and analytical have been proposed for analyzing mixed frequency distributions into their individual components. These methods include the use of probability paper (Harding 1949; Cassie 1954), graphical trial-and-error parabola fitting (Tanaka 1962), logarithmic differences (Bhattacharya 1967) and maximum likelihood (Hasselblad 1972). Several computer programs are available for applying these different methods to one or more samples: ELEFAN (Gayanilo et al. 1989), LFSA (Sparre et al. 1989), MIX (Macdonald and Green 1988, 1992), and MULTIFAN (Fournier et al. 1990). This insert presents Stata ado-files that apply Bhattacharya’s semi-graphical method (1967) for estimating the Gaussian components in size frequency data.
