near-bottom pelagic habitats of Long Island Sound. The analyses indicated 

 that the data base had a highly skewed distribution (due to a high 

 frequency of small catches and a low frequency of large catches) that 

 conformed to a negative binomial distribution. After appropriate trans- 

 formation, the data could be used in normal theory statistical tests 

 that, 80% of the time (beta level), could detect a 15% change with the 

 probability of 0.05 (alpha level) that a change that large occurred by 

 chance. In addition, significant autocorrelations existed in the data 

 bases, and these serial dependencies were found useful in time series 

 applications. 



