Dr. John Hough, Chief Science Officer, EPD International, Inc.

Every EPD has an associated accuracy value. Accuracies range from 0 to 1 and relate to the degree of reliability of an EPD estimate. All EPDs are estimates of an animal's true genetic worth. The EPDs with high accuracies are more dependable estimates of the animal's true genetic worth, while EPDs with low accuracies are less dependable estimates of its true genetic makeup.

How are EPD accuracies actually determined, though? Accuracy values are not some mysterious values drawn out of the sky! They are systematically
calculated for the Senepol Cattle Breeders Association through National Cattle Evaluation procedures conducted at the University of Georgia. Accuracy values are directly related to the amount of performance information available for an individual, its relatives and contemporaries.

Information used to calculate an animal's EPDs and accuracies include performance data from himself and his relatives such as parents, siblings, cousins and probably most importantly, progeny. Additional information used to determine accuracy values include the distribution of performance information in contemporary groups and the accuracy of an animal's contemporaries. With the use of multiple-trait evaluation procedures, performance information from one trait could also increase the accuracy of EPDs for other traits.

All EPDs and accuracies are calculated from performance information of animals relative to their contemporaries. A contemporary group is a group of cattle of approximately the same age and the same sex that have been given the same opportunity to perform. In order for an animal's performance record (weight or measure) to be counted, the animal must be raised in a legitimate contemporary group. If an animal is raised by himself with no contemporaries, his own weight cannot be used in calculating an EPD. Therefore, if an animal's own measurement can be used in his EPD calculation, his EPD accuracy will increase.

Parents, grand-parents, half-sibs, full-sibs, cousins and all other relative's performance information can be used in calculating an animal's EPDs.
Generally, the more performance information available from an animal's relatives, the greater is his accuracy. Typically, accuracy is higher when performance measures are available for both parents rather than just one. Having weights from 50 half-sibs will add to an animal's accuracy compared to weights for two half-sibs, and so on. But remember, just because an animal has lots of relatives does not mean his accuracies will be high. These relatives must have performance measures and they must be from legitimate contemporary groups.

Performance measures in some traits will increase the accuracy of EPDs in other traits because of multiple trait statistical analyses. The reason being that some traits are genetically related and therefore would help predict each other.
For example, weaning weight and yearling weight are genetically correlated. If a calf's weaning weight is very large compared to its contemporaries, we would expect his yearling weight would also be larger than his contemporaries because of the genetic correlation between weaning weight and yearling weight. Therefore, the accuracy of a bull's Yearling Weight EPD will be slightly higher if his weaning weight is used in the EPD calculations. The same sort of reasoning applies to weaning weight and milk as well, although a negative genetic correlation exists between growth and milk.

Progeny performance information has the potential of improving EPD accuracies to the greatest degree. A sire, though AI, can sire virtually an unlimited number of progeny, each of which could help contribute to his EPD accuracies. Table 1 shows the average number of progeny records associated with several accuracy levels. A sire's own progeny are relevant to birth, weaning and yearling weights while his daughter's progeny are pertinent to Maternal Milk accuracies. A bull's daughters' production is relative to Maternal Milk, because
milk production cannot be measured in bulls even though a bull transmits genes for milk production. Therefore a sire's genetic estimates for milk production must be derived from his daughters or other female relatives. Table 2 shows comparable information from a slightly different angle, the average accuracy values associated with different levels of progeny (or daughters for Maternal Milk). Average number of progeny vary between traits due predominately to differences in heritabilities for the different traits and differences in genetic correlations between the traits. Note that the accuracy levels and numbers of progeny or daughters are only averages. For example, the average number of progeny for cattle with a yearling weight accuracy between .25 and .34 is 2.9 progeny; although the range is actually between 1 and 11 progeny. You simply cannot say that a sire with seven progeny with birth weight records will have a Birth Weight Accuracy of .50! Other factors we have discussed also play an important role in the accuracy calculation. Distribution of progeny records also diplays an important role in accuracy determination. Four calf weights are not necessarily twice as meaningful as weights of two calves. Likewise, 50 records are not 50 times more accurate than one record. In calculating accuracy, the law of diminishing returns applies. A rough rule of thumb is the benefit of additional calf records starts to decrease with around seven to ten progeny in one contemporary group. Accuracy would be much greater for a bull with 50 progeny in ten contemporary groups with five progeny each, rather than having
all 50 progeny in only one contemporary group. Hopefully, this makes common sense that accuracy should increase with calves spread across multiple management and environmental conditions.

As you can see, there are many factors which are considered in calculating EPD accuracy values. Starting with the newly released 1992 Volume II Sire Summary, additional information for each bull will include total number of progeny with performance information, number of herds from which these progeny come and total number of daughters in production. This additional
information is not in lieu of accuracy values, but simply a way to provide more worthwhile information to Senepol customers.

Table 1.
Average Number of Progeny/Daughters By Accuracy Level
Accuracy
Range
Birth Weight
(Progeny)
Weaning Weight
(Progeny)
Maternal Milk
(Daughters)
Yearling Weight
(Progeny)
.25-.34 1.6 2.3 2.1 2.9
.35-.44 3.1 5.4 4.8 6.6
.45-.54 6.7 11.2 9.3 11.9
.55-.64 13.7 21.0 18.0 21.1
.65-.74 28.6 41.2 34.4 38.0
.75-.84 67.1 91.3 79.8 76.9
.85-.94 250.9 363.5 325.1 275.8
.95-.99 1575.1 2217.8 1775.5 1247.0
Table 2.
Average Accuracy Level By Number of Progeny/Daughters
Progeny* Birth
Weight
Weaning
Weight
Maternal
Milk
Yearling
Weight
1- 10 .42 .37 .30 .34
11- 20 .59 .53 .54 .51
21- 30 .66 .61 .62 .58
31- 40 .71 .65 .68 .64
41- 50 .74 .69 .71 .67
51- 60 .76 .71 .73 .70
61- 70 .78 .73 .76 .73
71- 80 .79 .75 .78 .74
81- 90 .81 .76 .79 .75
91-100 .81 .78 .80 .76
101-200 .85 .82 .84 .83
201 + .89 .88 .88 .91
*Daughters for Maternal Milk