nanometers, using a Coleman Junior II spectrophotometer. Concentration was 

 plotted against a Hycel (mammalian] standard curve. Red blood cell counts 

 were made at X 100, using an improved Newbauer hemacytometer. A modified 

 Hayme's solution (Heinle and Morgan, 1972) was the diluting medium for red 

 blood cell enumeration. 



Regression and correlation analyses were done according to Snedecor and 

 Cochran (1967). Simple and partial correlation coefficients were calculated 

 for each combination of parameters. Microhematocrit data were first trans- 

 formed to log to the base 10 (logj^g) 'to permit the use of parametric sta- 

 tistical procedures throughout the analyses. 



3. Results and Interpretation . 



Regression analyses between independent pairs of hematological param- 

 eters in the five fish species showed significant correlation between 

 microhematocrit and hemoglobin concentration in white perch, spot, and 

 striped bass (p < 0.01, Table B-1). Correlation between microhematocrit 

 values and red blood cell counts were also found to be significant in white 

 perch, spot, and hogchokers. Correlation of hemoglobin concentration and 

 red cell counts was significant in white perch and spot (p < 0.01). 



The significance of correlation is important in estimating a parameter 

 from a statistical relationship between it and another parameter. The 

 predictive capacity of the mathematical model is also important; e.g., the 

 correlation data for white perch. All three paired comparisons are sig- 

 nificant at the 0.01 level; i.e., chances are 1 in 100 (or less) that the 

 relationship established between any two of the blood parameters could be 

 due to chance alone. However, the correlation coefficients (r) differ by 

 as much as 0.21 between the microhematocrit with hemoglobin concentration 

 correlation (0.885) and the red blood cell count with hemoglobin concen- 

 tration correlation (0.676). These correlations may be significant at the 

 same probability levels, but the predictive capacity of the relationship 

 differs. This is shown by the coefficient of determination (r^), a meas- 

 ure of correlation which estimates the proportion of variance accounted 

 for by the correlation. Given a microhematocrit value with the hemoglobin 

 concentration correlation r^ = 0.784 from a white perch, the hemoglobin 

 content (y) is estimated by regression, knowing that the microhematocrit 

 value accounts for 78.4 percent of the variance in the hemoglabin concen- 

 tration estimate. 



In the estimation of red blood cell count from microhematocrit, 

 r^ = 0.464; only 46.4 percent of the variance of the predicted red blood 

 cell count can be accounted for by the microhematocrit. The correlation 

 coefficients were highly significant when estimating two blood parameters 

 from the same microhematocrit value, but the predictive capacity of the 

 former relationship was almost 80 percent; the latter was below 50 percent. 



The coefficients of determination must exceed 0.75 for a paired rela- 

 tionship before estimated parameters may be used for research purposes. 



63 



