Sample size and grouping of data for length-frequency analysis

Abstract A modified Bhattacharya method was used to analyze simulated length-frequency distributions in four different sample sizes (100, 200, 500 and 1000) and 20 interval sizes (2, 3, …, 20 and 50 mm). The relationships between optimum interval size, sample size, range and various biological characteristics were investigated. Life history characteristics ranging from a small, short-lived, fast-growing, high-mortality clupeid- or engraulid-type fish to a large, long-lived, slow-growing, relatively low-mortality lutjanid- or sparid-type fish were represented. Although the number of modes identified was generally small, results indicated that the best interval size for grouping the data was a function of sample size and biological parameters such as variability of length at age, recruitment pattern, growth rate and maximum size. Sample sizes > 1000 are required in order to identify more than half the modes in a typical distribution. These results suggested that life history characteristics should be taken into account when designing sampling strategies for length-frequency-based analyses. The proportion of modes identified can be increased by regrouping the data using a range of class interval sizes.