earlier studies distinct regional 

 effects of the St. Lucie Canal fresh- 

 water discharges. To study varia- 

 tions in catch-per-unit of effort 

 over time as correlated with dis- 

 charges, the typical stratified ran- 

 dom sampling plan over days, periods 

 and zones was waived in favor of 

 a systematic plan using a 4 x 4 Latin 

 square design to obtain weekly esti- 

 mates . 



SINGLE-SOURCE, CROSS CORRELATION 

 ANALYSES 



Single-source cross correlation 

 analyses, lagged up to 20 weeks, were 

 done to determine what effects the 

 following six inflow characteristics 

 had on fishing pressure: (1) week 

 rain (total weekly rainfall); 



(2) average St. Lucie daily flow/week; 



(3) average C-23 daily flow/week; 



(4) combined average C-23 and C-23A 

 daily flow/week. 



The results are summarized as 

 estimated lagged correlations. The 

 critical value for testing the sig- 

 nificance of the estimated cross cor- 

 relations was 0.324 for a 5 percent 

 test. 



MULTIPLE-SOURCE STEPWISE 

 REGRESSION ANALYSIS 



The primary purpose of this 

 study was to examine the effects (if 

 any) of St. Lucie Canal inflow on the 

 fish catch rates in the estuary. 

 Therefore, six characterizations of 

 freshwater inflow from the St. Lucie 

 Canal, as well as the two agricultur- 

 al canals, C-23 and C-23A, were 

 identified . 



A stepwise regression procedure 

 was utilized (SAS Procedure STEPWISE) 

 and adapted to selecting only that 

 subset of predictors whose assigned 

 regression weights are significantly 

 different from zero. The available 

 set of predictors consisted of 63 

 variables chosen from six weeks of 

 leading values taken from inflow 

 variables, with the level of sig- 

 nificance for retention in the model 

 set at 10 percent. This method has 

 a clear advantage of simulating a 

 situation in which inflows from C-23 

 and C-23A are held constant while the 

 relationship between St. Lucie inflow 

 and catch rate is being examined. 

 The coefficient of determination, 

 designated as R , represents an esti- 

 mate of the fraction of the total 

 variation in catch rate series which 

 can be accounted for by the respec- 

 tive series of the selected set of 

 predictors . 



FINAL MODEL SELECTION/RIDGE 

 REGRESSION ANALYSIS 



Using the results of the step- 

 wise regression analysis, the follow- 

 ing criteria were used for evaluating 

 the importance of the St. Lucie Canal 

 flow on species catch rates: (1) 

 there had to have been considerable 

 variation in the catch rate over 

 time for the species under considera- 

 tion; (2) the forcing of St. Lucie 

 variables on an existing model in 

 order to define a second model should 

 account for a considerable increase 

 in the fraction of the total variance 

 accounted for, 0.05 or more, as 

 reflected by the difference R - R„ ; 

 (3) essentially the same St. Lucie 

 variables should be selected when 

 presented initially as when forced 

 (as in the second model), otherwise 

 one would conclude that there was 



272 



