The terms reliability of prediction and accuracy of predcition in genomic selection are frequently used.
Plenty of papers are there discussing these points, however, I need the actual meaning of reliability.
Would you please explain what is reliability in predicting the Genomic Expected Breeding Value and how reliability differs from accuracy.
Generally, "accuracy" means "unbiasedness". An estimate is accuracte if it is correct, unbiased, if there is no systematic error. "Reliability", on the other hand, just expresses our expectation to get a similar estimate if we would repeat the experiment. Hence, "reliability" is related to "precision".
Generally, "accuracy" means "unbiasedness". An estimate is accuracte if it is correct, unbiased, if there is no systematic error. "Reliability", on the other hand, just expresses our expectation to get a similar estimate if we would repeat the experiment. Hence, "reliability" is related to "precision".
@J. WIlhelm, thanks for your answer. So it can be concluded that: Accuracy is unbiasedness and reliability means the reproducibility of the result under same condition.
Is it?
In many contexts in practical animal breeding, accuracy is the correlation between breeding value and estimated (predicted) breeding value, i.e. rAA(hat) and reliability is that figure squared: r^2AA(hat)
I think you get the answer from the above responses. In addition, the breeders prefer to use the reliability rather than accuracy, because it take into account the prediction error variance...
You can read more some nice explain in some papers
http://www.scribd.com/doc/109700016/Reliability-of-breeding-values-in-selected-populations
http://www.beef.org.nz/research/breeding/breed_acc.asp
or simply following some reviews from Prof. Ben Hayes or Prof. Theo Meuwissen which set up an idea for genomic selection 12 years ago:D.
A quote from Bijma 2012. Accuracies of estimated breeding values from ordinary genetic evaluations do not reflect the correlation between true and estimated breeding values in selected populations. JABG 129:345-358. ... "Thus, there are two different notions of 'accuracy'. For response to selection, the accuracy should reflect the correlation between true and EBVs in the candidates for selection, which is a property of a population, not of an individual. For the stability of EBVs, the accuracy should reflect the standard error of an EBV, which can be defined for a single individual..."
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics