Table 9,--Analysis of covariance of largest modal diameter 
on relative ovary weight regressions for the 
western and central equatorial Pacific 
Errors of estimate 
Degrees of |—— 
Source g Mean 
freedom 
squares 
Between regions 1 Sel) 
Within regions 35 879.430 
! whis unusually low figure is suspect, but recom- 
putation has shown it to be correct, 
Table 10, --Probabilities of ovaries occurring in various stages of maturity 
using two methods of classification 
Classification Classification by relative ovary weights 
Earl Lat 
by. modal Immature a y oKS. 
diameters maturing maturing Mature 
Immature 
Early maturing 
Late maturing 
Mature 
which it is placed is considered, this method of classification results in considerable error, 
especially in stages where the range of egg sizes is small, suchas the late maturing stage. 
Only 38 percent of the fish classified as late maturing would actually be in the late maturing 
stage, In the early maturing stage, however, which has a large range of egg sizes, 87 percent 
of the fish would be correctly classified, 
This error in classifying individual fish does not invalidate the method for use in group 
statistics, as the number of fish lost to one stage due to misclassification is somewhat balanced 
by the number of fish gained by that stage due to misclassification in the opposite direction; e.g., 
the number of immature fish classified as early maturing is somewhat balanced by the number of 
early maturing fish classified as immature. The calculated percentage of error mentioned 
earlier represents the error in the number placed in each group. The expected number ina 
group resulting from ciassification by relative ovary weights is either lower or higher than the 
expected number actually in the group. The amounts and the direction of error are 11.8 percent 
too low for the immature group, 0.3 percent too low for the early maturing group, 25.0 percent 
too high for the late maturing group, and 17.7 percent too low for the mature group. An example 
of the way these percentages were arrived at follows: the probabilities of classifying a fish in 
the immature stage according to relative ovary weights are 15 in a thousand (sum of column 
labelled immature, table 10); the probabilities of a fish actually being immature are 17 ina 
thousand (sum of row labelled immature); the error is 2/17 or 11.8 percent, 
Since the amount of material was too small for quantitative procedures in those aspects 
of the study which consider the stage of ripeness, the use of relative ovary weights as a measure 
of ripeness was thought to be justified. 
15 
