COMPARISON OF YELLOWFIN TUNA OF HAWAIIAN WATERS AND THE AMERICAN WEST COAST 



361 



The first dorsal spine was the longest on each of 

 the 188 specimens for wiiich this character was 

 measured. As noted on page 359, a linear regres- 

 sion did not provide a good fit to the original data, 

 compared with a linear regression fitted to the 

 logarithms of the variables. The latter is plotted 

 in figure 4. It was found that the same transfor- 

 mation apphed to the Costa Rican data, j-ielded a 

 linear regression with a shghtly unproved fit to 

 those data also (Schaefer 1948 fitted a Unear 

 regression to the original data) ; this regression 



TOTAL LENGTH mm. 



Figure 2. — Relations between length of second dorsal fin 

 and total length. Open circles represent Costa Rican 

 data; solid circles represent Hawaiian data. Solid 

 straight line is linear regression line fitted to Costa Rican 

 data. Broken straight line is linear regression line 

 fitted to Hawaiian data from fish 600 mm. and over in 

 total length. Solid curved line is second degree poly- 

 nomial fitted to all Hawaiian data. 



TOTAL LENGTH mm. 



Figure 3. — Relations between length of anal fin and total 

 L length. Open circles represent Costa Rican data; solid 

 p" circles represent Hawaiian data. Solid straight line is 

 i linear regression line fitted to Costa Rican data. Broken 

 [ straight line is hnear regression hne fitted to Hawaiian 

 j data from fish 600 mm. and over in total length. Solid 

 f curved line is second degree polynomial fitted to all 

 Hawaiian data. 



also is plotted in figure 4. Analysis of covariance 

 shows that the slopes of the two regressions do not 

 differ more than might be expected by chance, 

 but the levels do; the longest dorsal spines of 

 Hawaiian fish appear on the average to be a small, 

 constant percentage shorter than the longest 

 dorsal spines of Costa Rican fish. 



Similarly, the logarithms of length of longest 

 dorsal finlet against logarithm of total length 

 yielded a linear regression for the Hawaiian 

 measurements on all sizes of fish, and proved also 

 to provide a good fit to the Costa Rican data for 



