BAY OF BENGAL PROGRAMME Marine Fishery ResourcesManagement

INTRODUCTION Like most other biologists, fishery biologists often encounter situations where mixtures of distributions need to be separated into their component distributions. Such situations arise when, for example, the counts of different-sizedova in a fish gonad have to be separated into batches that willbe spawned successively,when length-frequencydata have to be separated into age groups or when "pulses" of recruitment~re to be quantified (Fig. 1L There exist a number of graphical and mathematical techniques which can be used for such separation (see Everitt and Hand, 1981, for a review). Most of them assume that the distributions to be separated from the mixture are and the ENORMSEP program of Yong and Skillman, do not require the external input of the expected number of component distributions in the mixture. The NORMSEP program, of which a ~ASIC version is presented here, also does require such an input. Thus, although NORMSEP is not the most sophisticated tool available to analyse size-frequency distributions, it is useful in situations where the data to be separated do not greatly overlap, or where previous knowledge exists on the number of probable positions of the means of the component distributions (see Mathew 1974). The version of NORMS~P presented here is a translation of the FORTRANIVversion of NORMSEP in Abrahamson (1971), written by V Hasselblad and modified by P K Tomlinson. The following changes were made to the original version : iii A routine was added which automatically computes, using the (guessed) means entered, a very large range of possible values for the standard deviation corresponding to each mean, by using a very small (0.01) and a very large (0.50) coefficient of variation (C. V. = s.d. V'X) to generate the range. This follows Abrahamson's suggestion that "anyone desiring to maintain the rigour of Hasselblad's original procedure only needs to place extremely wide bounds on the standard deviations". (ii) A routine was added which internally computes the number and approximate position of the cutoff points. Wi) The input routine has been made completely interactive, with prompts provided by the program, along with routines for checking and correcting the data entered from the keyboard. (iv) A routine was added along with prompts, to load length frequency data from a disk into memory, and use them as an input to NORMSEP. Such files must have been created by ELEFAN (see David et a/1982). No routine for checking the correctness of file inputs has …