The application of the log-linear model in the fishing industry

Abstract

This thesis is specifically related to the fishing industry. In the first part of this thesis, literature review is done on estimating the size-selectivity of gillnet. There are currently five widely-used methodologies about the size-selectivity of gillnet. After briefly describing all of them, respective examples about these five theories are given. The purpose of the first part is to clarify some useful statistical procedures for the problem encountered by fishing industries. -- In the second part of this thesis, the approach by Millar and Hoslt (1997) is employed and compared with the log-linear model. Using the data from fleet EW and fleet JV, it is found that the normal scale model yields the smallest deviance for JV, but not significantly smaller than that from the normal location model. For EW, the normal location model produces the smallest deviance, but not significantly smaller than the normal scale model. For both data sets, the Lognormal model yields the largest deviance.