Best Linear Unbiased Prediction

From BIF Guidelines Wiki
Revision as of 12:34, 28 October 2019 by Mnielsen (talk | contribs)

Best linear unbiased prediction (BLUP) describes properties of the mixed-model equations used for national cattle evaluation. Solutions to the mixed-model equations are best, in that they minimize differences between observed performance and performance predicted by linear combinations of the solved effects of factors affecting performance. The solutions are unbiased with respect to the available data, so that the expected difference between a predicted effect and an unknown true effect is zero. This does not mean that once a BLUP EPD is predicted for an animal, that a future EPD will never be more accurate. More data, particularly progeny performance, will result in a more accurate EPD than the EPD predicted before progeny were recorded. This also does not mean that unbiased BLUP can overcome biases that may exist in the data, statistical model, and simplifying assumptions used to generate the BLUP EPD.