696 



Fishery Bulletin 98(4) 



85 W 



Figure 1 



Estimated evasion index by set (estimated percentage of dolphins that evaded capture during 

 each set in relation to estimated initial herd size) for northeastern offshore spotted dolphins 

 ( 1992-95, Mexican fleet. PNAAPD data ). Three areas were delimited according to 60%, 50% , and 

 40% evasion index contours. 



are that data were measured in an intei-val scale (the test 

 requires at least an ordinal scale), and samples are consid- 

 ered to be independent when it is assumed that dolphins 

 remain approximately in the same area. Even though the 

 swimming capacity of dolphins is well known, it is diffi- 

 cult to establish how fast or far they are able to travel. 

 The Ki-uskal-Wallis test was preferred over the median 

 test (with simpler calculations) because the latter uses the 

 data more crudely than the former, and so the median test 

 will generally be somewhat less powerful (Conover, 1980). 

 The Kruskal-Wallis test ranks all observations from 1 to 

 n, and its statistic (//) is based on comparing each group's 

 mean rank with the mean of all the ranks, weighted 

 by the appropriate sample size to compensate for the 

 effect of unequal sample sizes (Neave and Worthington, 

 1988). If the differences among the three evasion areas 

 proved to be significant, nonparametric multiple compari- 

 sons could be executed to find which areas could be most 

 confidently claimed to have different medians from each 

 other (Conover, 1980; Neave and Worthington, 1988). The 

 statistic for multisample comparisons (T) is obtained by 

 calculating the absolute differences between the means of 

 ranks assigned to the samples and then dividing these dif- 

 ference by their standard deviations. The T statistic has 

 approximately the standard normal distribution; there- 



fore Hq will be rejected if T>2 (Conover, 1980; Neave and 

 Worthington, 19881. Statistical procedures were performed 

 with computing packages Statistica version 4.2 (StatSoft, 

 1993) and BMDP( Dixon, 1990). 



Differences between stocks To evaluate differences in 

 evasive behavior between eastern spinner and northeast- 

 ern offshore spotted dolphins, estimated evasion indices 

 by set for both stocks in evasion area 3 (as outlined in 

 Fig. 1 ) were compared. Areas 1 and 2 were excluded from 

 the analysis because eastern spinner sample sizes were 

 smaller than 30 and these sizes were considered insuf- 

 ficient for statistical analysis when compared to sample 

 sizes in area 3. Northeastern offshore spotted data were 

 not normally distributed, and nonparametric tests were 

 again used to compare the two independent samples (two 

 stocks). 



To decide which test was the most appropriate, "box- 

 and-whisker" plots ( Du Toit et al., 1986 ) were drawn to look 

 for general distribution similarities or differences between 

 both data sets i Fig. 2 ). The apparent difference in the medi- 

 ans would be worthwhile testing with the Mann-Wliitney 

 two-sample test (Conover, 1980; Neave and Worthington, 

 1988). However, data sets seemed to be different in spread 

 (Mann-Whitney assumes equal spread); therefore a more 



