FISHERY BULLETIN: VOL. 69, NO. 3 



among them suggestive of trends with tlie gen- 

 eral level of density. To define the trends, the 

 distributions for each specie^ group wei'e pooled 

 into three or four categories on the basis of the 

 extent of concentration in specific frequency 

 classes and the extent of dispersion over all 

 classes. Category designations are indicated in 

 the table, and percentage frequency histograms 

 are shown for each category in Figure 4. Num- 

 ber of blocks, number of night and day periods, 

 and the range of cruise medians (from Table 2) 

 are given for each histogram. 



The histograms suggest the same kind of trend 

 for the three crustacean groups: a shift from 

 distribution almost entirely restricted to the low^- 

 est classes when general area level is low, 

 through distribution of greater positive skew- 

 ness for intermediate levels of area density, to 

 symmetrical distril)ution as blocks of low me- 

 dian value disappear at the higher levels of area 

 density. The chaetognath distributions were 

 more difficult to classify, but it appears that fre- 

 quency distribution is appreciably skewed at all 

 levels of area density for this group. 



The differences in the trends for the four spe- 

 cies groups are also illustrated by the extent of 

 the overlap between the distributions for the 

 highest and lowest categories in the figui'es. 

 There is no overlap for small copepods, perhaps 

 50 to 75 'r overlap for large copepods and eu- 

 phausiids, and complete overlap for chaeto- 

 gnaths. These differences suggest that when the 

 overall area median is at one extreme, the possi- 

 bility of blocks with medians at the other ex- 

 treme is greatest for the chaetognaths, least for 

 the small copepods, and intermediate for the 

 large copepods and euphausiids. 



The existence of such trends in frequency dis- 

 tribution indicates that, at least for the crusta- 

 ceans, no one statistical distribution would satis- 

 factorily fit all the data sets; nor would any 

 single normalizing transformation be uniformly 

 effective for the different data sets. Without 

 normalized distributions, even the iiiterin-etation 

 of coefficients of variation would be diflicult in 

 comparing cruise periods. It may be noted, how- 

 ever, that when all frecjuency distribution cate- 



gories are pooled, the distributions for the four 

 species groups show similar degrees of skewness. 

 The total block array for each species group is 

 approximately normalized by log transformation. 



SMALL COPEPODS 



LARGE COPEPODS 



£^0^.000 [ riV^ 



C= 4 



P= 2 NIGHT 

 M=77-I0I 



CATEGORIES POOLED 

 0=87 



3 6 9 12 15 18 21 24 

 NUMBER/m' X lO' 



EUPHAUSIIDS 



rm^" 



c- 3 



n-20 



P= 2 NIGHT 

 41-43 



lOOr 



C= 2 

 n=26 



P= 3 NIGHT 

 M= 12-23 



n:4I 

 P= 5 DAY 

 M= 1-13 



CATEGORIES POOLED 



25 50 75 100 125 150 175 



number/m' 



n 



XI 



P= I DAY, I NIGHT 

 M=42-50 



P = 2 DAY. 3 NIGHT 

 M= 5.000-6000 



P= 2 NIGHT 

 M=2l-30 



n-. 29 



P=2DAY, I NIGHT 

 M^BELOW 3.000 50- 



CATEGORIES POOLED 

 n=e7 



50 100 150 200 



NUMBER/M* 



CHAETOGNATHS 



C= 3 



n= 24 



P: 2 DAY, I NIGHT 



M-57-I2I 



P= 2 DAY, 3 NIGHT 

 M= 12-36 



P: I DAY, I NtGHT 

 M= 21-33 



CATEGORIES POOLED 

 n=e7 



50 100 150 200250300 



number/m' 



Figure 4. — Percent frequency distribution histograms 

 for four species groups as pooled on the basis of simi- 

 larity. Range of sample medians (M) (from Table 2), 

 number of day and night periods (P), and number of 

 sampling blocks (n) represented by each histogram 

 category (C) is shown. The histogram at the bottom 

 of each set shows the distribution of all block medians 

 from all cruises. 



688 



