Interpreting EPD Accuracy
Livestock Update, March 2002
Scott P. Greiner, Extension Animal Scientist, Beef, VA Tech
The embrace of Expected Progeny Differences by beef producers has allowed for rapid, predictable genetic progress in the beef industry. Proper use and application of EPDs requires an understanding of what the EPD values represent and what they do not. Accuracy values become very relevant in this context, as they are a measure of possible change or "risk" associated with an EPD. Put another way, accuracy values are indicators of the reliability of the published genetic estimates for an animal.
Accuracy values are published for all EPD values reported on an animal. Accuracy can be defined as the relationship between the estimated EPD of the animal and the "true" EPD of the animal. This relationship is expressed numerically from zero to one. As the accuracy value approaches 1.0, the EPD reported is more likely to represent the true genetic merit of the animal. Conversely, low accuracy values (closer to zero) indicate that the reported EPD is less reliable. Accuracy is primarily a function of the amount of information available to calculate an EPD for any given trait. Information, primarily in the form of performance records, is derived from several sources to estimate EPDs on a given animal. These sources include records on the animal itself, its sire and dam, collateral relatives, and progeny records. As the volume and quality of records used in the estimation of an EPD increases, so does the confidence we have that the EPD has been estimated correctly (accuracy).
The following table will be utilized to explain the implications of accuracy.
|Birth Weight EPD||Accuracy||Possible Change||"true" EPD Range|
|Bull A||+2.0||.25||±2.4||-0.4 to +4.4|
|Bull B||+2.0||.90||±1.2||+0.8 to +3.2|
Bull A and B have identical BW EPDs, but differ considerably in their accuracy values. Bull A would be typical of a yearling bull, with his EPD derived from pedigree information and his own individual performance. Most yearling bulls will have accuracy values ranging from .10 to .35 for growth traits. Bull B would be typical of a sire with a large number of progeny, and likely used AI across several herds. A practical way to evaluate accuracy is to put it in the context of associated possible change. Possible change defines how much we might expect the current EPD to change (plus or minus) as more information is collected and used in the estimation of the EPD. For Bull A, an accuracy value of .25 for BW EPD is associated with a possible change of +2.4 pounds. Therefore, we would expect his "true" BW EPD to be between -0.4 and +4.4 pounds (2.0+2.4) 68% of the time. Bull B, with a higher accuracy value, has a much lower possible change (+1.2) and therefore smaller range that we expect his true EPD to fall within (+0.8 to +3.2). It is important to recognize that EPDs are our best estimates of an animal's genetic worth. Due to a variety of potential sources of error, we never know the "true" EPD for any trait on any animal. Accuracy values, therefore, indicate how much we know about the animal's true genetic worth and how confident we can be in the estimated EPD.
Implications of EPD accuracy deal with associated risk. If the two bulls previously discussed were being considered for use on heifers, there would be much lower risk associated with Bull B. Due to the high accuracy value of Bull B, it is less likely that his EPD will change with additional information. Even if it does change, it is unlikely that his BW EPD will go up (or down) significantly. Comparatively, Bull A has a larger possible change and there is more risk that his EPD could change with additional information (the primary risk would be that his BW EPD become substantially heavier than estimated). This example illustrates a primary advantage of using high accuracy, low BW EPD sires through AI on heifers. Similar examples can be given for all EPD traits, and possible change values can be found in sire summaries of all breeds.
An important concept to understand is that EPDs, regardless of accuracy, are our most powerful tool to make genetic change in beef cattle. EPDs have been estimated to be over seven times more reliable than adjusted weight records, ratios, and visual appraisal. Even on young bulls with relatively low accuracy values, EPDs are our most objective indicator of the animal's genetic merit. For all practical purposes, high accuracy sires are available only through AI. Therefore, most of natural service bull-buying decisions will be made using relatively low to moderate accuracy EPDs. Keep in mind when evaluating possible change that there is an equal chance that an EPD will go higher as opposed to go lower (or get "better" vs. "worse"). When evaluating young bulls, small differences in WW and YW EPD become less significant due to accuracy and possible change (large overlap in the range of their "true" EPDs).
A common misconception is that accuracy is an indicator of expected variation in a resulting calf crop. Accuracy and possible change are not related in any way to progeny variation. High accuracy EPD animals would not be expected to have any more or any less variation in their calf crop compared to low accuracy EPD animals.
In summary, accuracy values assist us in more thoroughly understanding an EPD. They are an important risk management tool, to be used in conjunction with an EPD and other important selection criteria.