A mixed model analysis of a selection experiment with Merino sheep in an arid environment
Erasmus, Gert Johannes
MetadataShow full item record
1. A mixed model analysis of a selection experiment with Merino sheep carried out at the Klerefontein Research Station, Carnarvon, C.P., from 1962 to 1982 was undertaken. The experiment consisted of three selection lines: One objectively selected for high clean fleece mass, another selected subjectively for "overall excellence" and an unselected control line. The generation interval was almost identical (3.3 years) in all three lines and the experiment represented roughly six and a quarter generations of selection. 2. Heritability estimates of 0.247 for body mass, 0.229 for clean fleece mass and 0.369 for mean fibre diameter were obtained using Henderson's Method 3. REML estimates on a sire model yielded almost identical estimates. 3. The line selected for high clean fleece mass showed a genetic response of 0.014 kg per annum in this trait. The low response is ascribed to a low heritability and a low cumulative selection differential due to low reproduction rates. The "realised heritability" was 0.187. Small annual genetic changes of 0.098 kg and 0.016 micrometer in body mass and mean fibre diameter, leading to "realised heritabilities" of 0.205 and 0.282 respectively, also occurred. 4. The largest response in the subjectively selected line was for body mass (0.189) and fibre diameter (0.042), while clean fleece mass increased by 0.008 kg per annum. "Realised heritabilities" obtained were 0.234 for body mass, 0.138 for clean fleece mass and 0.401 for fibre diameter. 5. The control line remained genetically stable for clean fleece mass and fibre diameter, but showed a significant response of 0.168 kg per annum for body mass. The "realised heritability" obtained was 0.210. This response is ascribed partly to natural selection in a harsh environment where some sheep were at a sub-optimum body mass. 6. An illustration of how lower heritability values in the mixed model equations leads to smaller predicted gains and hence lower "realised heritabilities" and also lower variances of predicted breeding values is provided. 7. The average inbreeding coefficient was roughly 0.02 at the end of the experiment and no significant differences in inbreeding among the three lines existed. 8. It is concluded that a mixed model analysis of selection experiments provides useful information especially if there is reason to believe that the control is not genetically stable.