Comparison ol variance component estimators in twa statisucal-genetic methods of crosses.
A numeric comparison was made, of the variance components estimators, constant adjustments, maximum likelihood and minimum variance quadratic unbiased estimator (MIVQUE), for the models proposed by Comstock & Robinson (1948, 1952). They are equivalents to lhe models of random hierarchic classification, and mixed factorial with interaction, using as the main comparison criterium the mean squared error. The data were simulated from the normal distribution, for each model, considering two variance relations, a?/a2 where ai is the variance of the random effect, but not the error, and a 2 is lhe error variance, taking for each case the balanced and unbalanced datas, with 5% of empty units (cells). Among the three methods, the MIVQUE will be preferable, if the computacional resources are easily available, because of its speed and efficiency as estimator. Even though, uniformily, between this one and the constant adjustment method, one is not better than the other for the models studied.
| Main Authors: | , , |
|---|---|
| Format: | Digital revista |
| Language: | por |
| Published: |
Pesquisa Agropecuaria Brasileira
2014
|
| Online Access: | https://seer.sct.embrapa.br/index.php/pab/article/view/13337 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A numeric comparison was made, of the variance components estimators, constant adjustments, maximum likelihood and minimum variance quadratic unbiased estimator (MIVQUE), for the models proposed by Comstock & Robinson (1948, 1952). They are equivalents to lhe models of random hierarchic classification, and mixed factorial with interaction, using as the main comparison criterium the mean squared error. The data were simulated from the normal distribution, for each model, considering two variance relations, a?/a2 where ai is the variance of the random effect, but not the error, and a 2 is lhe error variance, taking for each case the balanced and unbalanced datas, with 5% of empty units (cells). Among the three methods, the MIVQUE will be preferable, if the computacional resources are easily available, because of its speed and efficiency as estimator. Even though, uniformily, between this one and the constant adjustment method, one is not better than the other for the models studied. |
|---|