A comparison of exact and sequential methods in multi-stage index selection

Area of Science:

• Animal Breeding and Genetics
• Quantitative Genetics
• Statistical Genetics

Background:

• Sequential multi-stage index selection is a common method in quantitative genetics.
• The existing theory often assumes zero correlation between indices at different selection stages.
• This assumption can lead to inaccuracies in estimating genetic gain and selection proportions.

Purpose of the Study:

• To evaluate the impact of non-zero correlation between indices on the accuracy of sequential selection methods.
• To correct estimation errors in genetic gain and selection proportion caused by the zero-correlation assumption.
• To analyze the influence of selection intensity and inter-stage index correlation on the sequential method's performance.

Main Methods:

• Utilized knowledge of means and volumes of truncated multivariate normal distributions for corrections.
• Analyzed the effects of selection intensity and the correlation coefficient (ϱ) between first and second stage indices.
• Focused computational analysis on two-stage index selection procedures.

Main Results:

• The sequential method showed good performance when the correlation (ϱ) was below 0.6.
• Accuracy rapidly declined as the correlation (ϱ) exceeded 0.6.
• Selection intensity had a lesser impact on accuracy compared to index correlation (ϱ).
• Errors in selection percentage and underestimation of genetic gain increased with selection intensity, while overestimation decreased.

Conclusions:

• The implicit zero-correlation assumption in sequential index selection theory can lead to significant errors.
• Corrections using truncated multivariate normal distributions improve the accuracy of genetic gain and selection proportion estimates.
• The correlation between indices is a critical factor, with accuracy deteriorating sharply above 0.6.
• Selection intensity also affects accuracy, but to a lesser extent than index correlation.