The amount of feed applied to the cages in the C ponds was about 1.5 times as high as that applied to the B ponds,due to the faster growth of fish in the (laeger) C ponds.
Fish in all cages were weighed every 2 weeks,in order to estimates their mean weight and compute the feed ration.
The fry were stocked on July 23 and the test was terminated after 143 days(on december 13)
Experimemtal plan
The four groups produced as described above (i.e. Large jumpers, laggards ,
Small jumper and small laggards ) were stocked into cages either separately , at 6 fish of the same group per cages, or communally with 3 fish of each of two groups.
The aim of this of this stocking scheme was to compare the effect of grading and mutiple nursing under conditions of competition between the test groups,and in the absence of such competition.
Samples of each genetic line were stocked into) cages in each pond .
Cages standing in ponds c-23 and b-1 were stocked with one replication of each line,and those in ponds c-24 and b-2 with the other replication.
To evaluate the effect of grading,each unit of 6 cages included one cage stocked separately with jumpers,one separately with laggards and a further cage stocked communally with jumpers and laggards.
Similarly for mutiple nursing,for each line one cage was stocked separately with large fingerlings, one with small fingerlings and a future cage communally with large and small ones .
The stocking scheme is described in Table 1. Note that the groups (large jumpers +large laggards ),which test the effect of grading under conditions of competition and large (jumpers +laggards ), testing the effect of multiple nursing in the absence of competition when compared to small(jumpers +laggards ),are replications of the same combination.
Similarly (small jumpers +small laggards ) and small(jumpers +laggards ) are replications.
120
Methods of estimating the association between weight gain and initial weight.
In our genetic testing program,carried out in ponds and cages,pairs of multiple nursed samples were produced from a number of test progenies.
The association between weight gain and initial weight,used as a correction factor, was estimated by dividing the mean difference in weight (d)
,i.e.the correction factor = D/d .
In the present investigation,the association between weight gain difference and initial weight difference were computed from a regression equations D=a+bd.
In this situation the correction factor D/d = a/d+b .
The correction factor thus consists of two terms,the regression coefficient of D on d , plus a factor computed by dividing the y-intercept of this regression line by the differences in mean initial weight between the weight groups (d).
This factor is true correction factor only when weight variation between samples was generated by multiple nursing.
In the cage of graded weight classes,it is incorrect to use it for correcting the bias generated by initial weight variation among genetic groups,since weight variation among greded weight classes is due to phenotypic and not purrly environment differences.
These "pseudo-correction factors" were compared with true correction factors in order to compare grading with mutiple nursing.
124
Comparison of the effects of mean initial weight difference generated by grading and multiple nursing on mean weight gain.
The data presented in Tables 2 and 3 were used to estimate the association between mean initial weight difference and differences in mean weight gain.
These relations were estimated for grading and multiple nursing and for separate and communal testing.
125
The association between grading and multiple nursing in separate testing are shown in fig 3.
The figure shows that the coefficient of regression (b) of differences in mean weight gain (D) on difference in mean initial weight (d) and the y-intercept (a) are both higher for grading than for mutiple nursing.
This implies that the expected difference in mean initial weight (d),and increases with increasing values of d.
In communal testing,on the other hand, the coefficient of regression of mean weight gain on mean initial weight is higher for mutiple nursing than for grading,whereas the y-intercept is higher d-values.
The factors (a/d+b) for the difference treatments, i.e. Grading and multiple nursing in separate and communal testing,are shown in Table 5 (last column ).
Inspection of the table shows that,under conditions of separate testing,the factor is higher for grading than for mutiple nursing (2.27 vs 1.54)
.
The same tendency also appears in communal testing,but the difference is smaller (3.3 vs. 3.18).
Comparison of the influence of mean initial weight variation in communal and separate testing.
The association between difference in mean initial weight and mean weight gain show that when initial weight variation was generated by multiple nursing, the coefficient of regression (b) and the y- intercept (a) are both larger in communal than in separate testing.
As a result,the correction factor (a/d+b) for mutiple nursing is twice as high for communal than separate testing (3.18 vs. 1.54 ) (see Table 5 and fig 4).
When mean initial weight difference were generated by grading,the coefficient of regression of mean weight gain on mean initial weight is higher for separate than for communal testing.
126
The factor (a/d+b) is higher for communal testing (3.33 vs 2.27) due to the higher y-intercept (34.3 vs. -0.5) (see Table 5)