FISHERY BULLETIN; VOL. 87, NO. 3, 1989 



sion and avoidance by dividing each A'^,, by the 

 estimated value of P^ for the appropriate length 

 interval. To better visualize the effect of this 

 correction, we chose A^,, from a data set that was 

 considerably larger (A^ = 4,178) than those used 

 in the net comparisons because histogi'ams of 

 this larger data set were smoother in appear- 

 ance. This larger data set comprises all larvae 

 that we have measured to date (including those 

 obtained in the net comparisons) and that were 

 collected during the day by the standard net. 



The variance of the estimated value of A^ was 

 approximated by using bootstrapping (Efron and 

 Gong 1983): 1) each of the four experimental 

 length-frequency data sets was randomly sub- 

 sampled with replacement to produce four new 

 samples with the same sample sizes as the orig- 

 inals; 2) Equations (5) and (6) were fit to the four 

 synthetic samples by the methods described 

 above; 3) Pes, Per, Pis, Pre, ^nd Pc were esti- 

 mated for each length interval; and 4) A^ for each 

 length interval was estimated by dividing the A^„ 

 from the large sample of nehu length-frequency 

 data by the estimated value of P,.. This pro- 

 cedure was repeated 500 times, generating 500 

 independent estimates of Pc for each length. To 

 reduce variance owing to rare but extremely 



large estimates of A'^ produced when P^ was near 

 0, the data were trimmed by eliminating the 25 

 largest estimates within each length interval (5% 

 of the sample). After data trimming, the vari- 

 ance was calculated among the remaining 475 

 independent estimates. 



The variances of Pes, Per, Prs, and P,e were 

 also calculated from the same 500 independent 

 estimates (no data trimming was required). A 

 two-sample t-test incorporating these variances 

 was then used to test for significant differences 

 between Pes and Pe,- and between P,s and P,.^ 

 within each length interval. 



RESULTS AND DISCUSSION 



The estimated entry probabilities for the stan- 

 dard net (Pes) decreased from 1.00 for 3 mm 

 larvae to near 0.00 for 10 mm larvae, whereas 

 the entry probabilities for the small mesh net 

 (Per) decreased from 0.95 for 3 mm larvae to near 

 0.00 for 8 mm larvae (Fig. 1). When the apparent 

 difference in entry probabilities between the two 

 nets was examined statistically for each 0.5 mm 

 length interval between 2.5 and 10.75 mm, Pe,- 

 was found to be significantly less than Pes (two- 

 sample, one-tailed f-test; P < 0.05) for all length 



0.00 



2.0 4.0 6.0 



8.0 10.0 12.0 14.0 

 LENGTH (mm) 



16.0 18.0 20.0 



Figure 1. — Entry probabilities for the standard (Pes) and test nets (Pe,-) and reten- 

 tion probabilities for the standard (P^s) and test nets (P^e) are shown by 0.5 mm 

 length intervals (upper panel). Capture probability (Pe) for the standard net is 

 shown by 0.5 mm length interval (lower panel). 



450 



