time-series analysis, the Krumbein and Tukey (1956) method of analyzing 



proportions, and non-parametric tests. 



A variety of techniques were used on the seine catch data to elucidate 



the underlying probability distribtuion. The frequency distributions of 



the CPUE and log (CPUE+1) transform were evaluated visually. The 



Kolmogorov-Smirnov D- statistic was used to test for normality. The 



possibility that the data came from a negative binomial distribution was 



investigated. To do this the mean, m and a parameter, k, must be known 



(Poole 1974). The mean can be estimated by the sample mean x, while k 



may be estimated in one of three ways (Bliss and Fisher 1953) . The 



simplest solution for k is: x 



k 2 = (1) 



where x = sample mean 



s = sample variance 



This estimator of k is 90% efficient for small values of the mean when 

 (k/x) >6, for large values of x when k >13 and for intermediate values 

 of x when ( (k+x) (k+2) /x)^ 15. 



An alternate method requires that estimates of k are iteratively 

 replaced in the following equation until the two sides balance: 



k * log e (1 + (x/k)) = log e (N/f ) (2) 



where N = total number of samples 

 f Q = number of zeros 



This estimator if 90% efficient if (f /N) > 1/3 and x < 10. For x > 10, 



o 



((x+0.17)(f /N)-0.32) should be greater than 0.20. 

 o 



41 



