Prediction of Lactation Yield of a cow from Her Partial Yield

Abstract

Investigation, using data from seven Indian herds, has been carried out on the problem of predicting lactation yield of a cow from her partial yield. 1. Correlation coefficients were found between lactation yields and partial yields and also between partial yields and inter yields. In the four herds, namely, Red Sindhi, Hosur, Red Sindhi, Bangalore, Tharparkar (weaned) and Haryana, I.V.R.I., sixteen weeks’ yield, with a correlation of the order of +.90 with lactation yield, was found to be a good predictor. This initial period covered about 50 per cent, of the lactation yield and about 35 per cent. Of the lactation length. In the case of cost of milk production survey, Delhi, an initial period of twenty weeks, covering about 60 per cent. Of the lactation yield and about 50 per cent of the lactation length, was found sufficient to attain the desired correlation of +.90. For the remaining two herds of Haryana (Hissar) and Tharparkar (not weaned) cows, the required period was as long as twenty eight weeks; this period covered about 60 per cent. Of the lactation length and about the same (60) percentage of the lactation yield.

2. It was observed that the power of prediction was poor in the case of a single cow’s yield, the standard error of the predictor varying from about 10 per cent. To thirty per cent. However, prediction could be made fairly accurately in the case of average yield of a given group of even a for cows such as ten. It was further observed that the percentage standard error of prediction was much lower for high yielding cows than for 100 yielding ones. Prediction line and its 85 per cent. Fiducial bands have also been graphically represented.

3. Use of curvilinear regression was soon to be of only very slight help in improving the production. On the other hand the proportion of variation explained by the linear prediction formula could be increased more by extending the period of initial yield.

4. About ten daughters were found to be necessary to establish the significance of the merit of a bull which is twenty per cent. Superior to a given herd average. The use of partial records of the daughters would increase this number to about a dozen. For establishing ten per cent. Superiority. However, as many as thirty five daughters were found necessary, and this number would be increased by a large extent in the case of partial records.

5. The problem of predicting lactation yield from partial yield which itself is estimated from a sample of yields on randomly selected days of the initial period was also investigated. Systematic sampling at weekly intervals and equivalent simple random sampling were considered. The former procedure was found to be more than eight times as efficient as is the latter technique in estimating the partial yield. It was found, in addition, that loss of efficiency because of the sampling of partial yields was less than four per cent. In the case of a single cow’s yield while for a group of ten the loss was only of the order of ten per cent., the sampling technique being the ‘Systematic sampling’. It was thus found that, at first, partial yield could be estimated from an appropriate sample and this estimated yield could then be used, with reasonable safety, in the prediction equation to arrive at, finally, the lactation yield.