Date of Award
Doctor of Philosophy
Records of birth weight, average daily gain and 120- day weight of 447 purebred Hampshire lambs in 19 sire groups born from 1952 to 1968 at the Main Experiment Station of The University of Tennessee Agricultural Experiment Station were used to calculate estimates of heritability before and after least-squares adjustments of the records for the significant effects of age of dam, month of birth, sex of lamb, type of birth, type of rearing and year of birth. Estimates from unadjusted data were erratic and inexplicable, whereas, those from adjusted data were reasonable and consistently positive.
Estimates of heritability from regression of off spring on dam, from regression of offspring on mid-parent and from paternal half-sib correlation were very similar in magnitude, but estimates from regression of offspring on sire were much larger than the other regression estimates and had much larger standard errors reflecting larger sampling errors due to the small number of sires.
The heritability estimates ranged from 0.016 ± .038 to 0.388 ± .180 for birth weight, 0.101 ± .028 to 0.330 ± .320 for average daily gain and 0.099 ± .038 to 0.218 ± .261 for 120-day weight, the largest estimates being those obtained from the regression of offspring on sire. There were no appreciable differences among the estimates obtained from regression of offspring on dam, regression of offspring on mid-parent and from paternal half-sib correlations, but the estimates obtained from regression on sire were much larger and had much larger standard errors than those obtained by other methods. It was concluded that heritability estimates based on regression of offspring on sire cannot be as accurate as estimates obtained by the other methods because of the much larger sampling errors to which they are subject as a result of small numbers of sires.
In view of non-significance of differences between estimates obtained using individual offspring values and those obtained using progeny means in the parent-offspring methods and some statistical advantages of using progeny means over using individual offspring values, it was concluded that use of progeny means is slightly preferable to use of individual offspring values.
There was no significant indication of the existence of variance due to maternal differences with respect to the three traits studied here as indicated by the lack of significance between paternal half-sib estimates of heritability, which receive no contribution from maternal variance, and parent-offspring estimates which contain a contribution from whatever maternal variance exists.
Extremely large estimates of genetic correlations were obtained (ranging from 0.76 to 0.98), and it was concluded that these are gross overestimates of the respective parameters due to extremely large sampling errors to which estimates of genetic correlation are subject in data from samples of animals numbering many fewer than 1000 head.
Srinivasan, G., "A comparison of heritability estimates obtained by different methods. " PhD diss., University of Tennessee, 1969.