National Swine Improvement Federation

            Purdue University Cooperative Extension Service
                        West Lafayette, Indiana

Fact Sheet Number 13                                      NSIF-FS13

                Across-Herd Genetic Evaluation of Swine

              C. M. Wood, Virginia Polytechnic Institute;
                  T. S. Stewart, Purdue University;
                 J. W. Mabry, University of Georgia.

            Larry Young, USDA MARC, Clay Center, Nebraska;
                   Bruce Leman, Roanoake, Illinois;
            J. R. Jones, North Carolina State University;
              Darrel Anderson, West Lafayette, Indiana;
                  Ron Bates, University of Missouri.


      What is across-herd genetic evaluation? Across-herd genetic
evaluation is the comparison of sires, dams, and/or pigs in different
herds based on the analysis of performance information (such as
backfat) collected on pigs of the same breed in different herds and at
many different times. It may include on-farm records, test station
data, national breed test information, and progeny tests. Such
evaluations can be used to accurately compare animals at different

      Across-herd genetic evaluations are used routinely by the dairy
and beef industries in the United States to identify the very best
genetic material in each breed. Canadian pork producers use an
across-herd program to select breeding stock. In the U. S., the STAGES
program began using an across-herd evaluation for Yorkshire data in
1990, and the Hampshire, Yorkshire, Spot, and Duroc breeds use a
similar program to evaluate centrally tested boars across the country.

      Why use across-herd genetic evaluation? A major goal of
seedstock breeders is to increase the rate of genetic progress to
better meet the needs of commercial operations producing high-quality
pork for consumers.  Accurate, timely information is vital in making
selection decisions to meet that need. Across-herd genetic evaluations
offer several advantages over other evaluation methods.

1.    All available performance information about an animal and its
      relatives (parents, sibs, offspring, etc.)  is used to obtain
      the most accurate estimate possible about the genetic merit of
      an animal. These estimates of genetic merit are reported as
      Expected Progeny Differences (EPDs), which predict how
      off-spring of an animal will perform when compared to average
      animals within a breed (NSIF - FS8).

            Because EPDs are expressed as deviations, they may be
      positive or negative, and the breed average is zero. This
      differs somewhat from within-herd evaluations, in which the herd
      average is calculated and half the animals in that herd are
      above the herd average and half are below average. Based on an
      across-herd evaluation, all animals within a herd could be
      better than the average of the breed, but the converse also is
      true.  In practical terms, neither extreme is very likely; the
      vast majority of herds will have some animals above the breed
      average, and some below. An additional benefit of using EPDs is
      that the unit used for reporting EPDs is the same as that used
      for the trait, such as pigs for number born alive or pounds for
      average daily gain.

2.    Because information comes from many herds, fair comparison of
      animals in different locations is possible with across-herd
      evaluations. In contrast, comparing raw data on animals
      evaluated within herds (or test stations) to others in different
      locations is tempting, but such comparisons may result in poor
      selection decisions because performance differences between pigs
      in different locations may be environmental and/or genetic.

3.    Use of EPDs from across-herd evaluations to select breeding
      animals will speed up genetic change within a breed, because the
      best of many animals will be chosen as parents of the next
      generation.  A large data base consisting of many animals
      alleviates some of the negative influence that low
      heritabilities have on genetic progress in traits such as litter

4.    Across-herd evaluations offer the opportunity to effectively test
      for traits like carcass merit that are difficult to evaluate on
      a within-herd basis. For example, because carcass traits require
      slaughter of animals, fewer records are collected, and
      across-herd evaluations can make the best use of data that are

Genetic Principles of Across-Herd Evaluations

      Across-herd genetic evaluations employ the same genetic
principles as within-herd evaluations, extended to include
between-herd differences. The goal is to increase the rate of genetic
progress in traits of economic importance to the industry. What
follows is information on aspects specific to across-herd evaluations.
Detailed information about applications of genetic principles may be
found in NSIF - FS9.

      Genetic base. The genetic base consists of the initial animals
in the population to be evaluated. Across-herd evaluation programs
usually define a population as animals of the same breed, because
individuals share many of the same genes, making predictions (EPDs)
more consistent. The genetic base is frequently defined as all animals
in all herds during a specific time period, such as the first five
years of a testing program.  Specific definitions may vary from breed
to breed. By broadening the population to the entire breed, more
animals are available from which to choose the next generation. This
results in a greater selection intensity, and hence enhanced genetic
progress. Environmental influences which vary from herd to herd over
time can be accounted for during the evaluation process by comparing
performance to the base period. This allows a com- parison of animals
based predominantly on genetic merit. Such procedures also allow
estimation of genetic trends, or changes in the average breed genotype
over time, compared to the original (base) animals. When genetic
progress is occurring, later generations will be genetically improved,
but estimated breeding values on older animals will be biased unless
all animals are adjusted relative to the genetic base.

      Contemporary groups. The importance of the proper definition of
contemporary groups by the breeder cannot be overemphasized, because
contemporary groups form the basis of all evaluation programs. NSIF
Factsheet 5 (Per- formance Records for Selection Programs) has a good
discussion of the concept of contemporary groups. It is important to
avoid single-sire groups, but producers also need to balance size of
each group with uniformity within the groups. A good rule of thumb is
to include offspring of at least three sires in each group, prefer-
ably 40 to 50 per sire. In general, the more progeny per sire the
better, as long as they can be managed uni- formly.


Figure 1. Connecting Herds with Genetic Links.

                      |  TEST STATION  |
                      |                |
                      | B1 B2   B3 B4  |
                        .  .      .  .
              ----------.-.        .  .-----------
             |  HERD 1  .. |        . |.  HERD 2  |
             |          .  |         .| .         |
             | S1  S2  S3  |          ..S4     S5 |
              ------.------            -----------
                      .   -----------
                        .|   HERD 3  |    --------------
                         . .         |   |   HERD 4     |
                         |  ..B5     |   |              |
                         |           |   | S9  S10  S11 |
                         | S6 S7 S8  |    --------------

Herd 1 is directly linked to Herd 3 and the test station through sons
of S2 (B5) and S3 (B1 and B2), respectively. Herd 2 is directly linked
to the test station through sons of S4 (B3 and B4). Indirectly, Herd 1
is linked to Herd 2 through the boars in the test station; Herd 3 is
linked to the test station through Herd 1, and to Herd 2 through the
test station linked to Herd 1. Herd 4, however, is not linked at all
because of the lack of genetic ties. In this scenario, across-herd
evaluation allows comparison of Sires 1-8 and Boars 1-5, but not Sires

      Connectedness. One major advantage of across-herd evaluations is
that they tie together information from many sources by using genetic
(pedigree) relationships among animals being evaluated. As shown in
Figure 1, once a genetic link is established between groups of
animals, all animals in those groups can be compared, whether they are
related or not. However, just as a rope is made stronger by braiding
strands together, the more genetic links (connectedness) between
groups, the more accurate the EPDs will be.  As a result, it is impor-
tant that enough links be present to make an across-herd evaluation
worth the effort.

      For those herds participating in earlier phases of STAGES,
within-herd evaluations made over time use genetic links to connect
contemporary groups within a herd, and many herds within a breed are
already linked by exchange of breeding stock as well as through
artificial insemination (AI) and embryo transfer. Test stations can
contribute genetic links by evaluating animals from different herds at
the same location, and by sending littermates or half-sibs from one
herd to several stations (see NSIF Factsheet 11). The fastest way to
strengthen genetic ties, however, is through AI, which can also make
superior sires available to more produc- ers, and help maintain a
broad genetic base.

      Accuracy. Because EPDs for the same trait on the same animal can
change over time, depending on the amount of information available,
and how well progeny actually perform, accuracies are assigned to help
producers choose among animals that have similar EPDs. The accuracy is
a risk management factor: the closer it is to 1.0, the more likely
that the 4 average performance of offspring will be close to the
parent's EPD, and the less likely an animal's EPD will change
drastically with the addition of that new information. Thus, the more
performance tested progeny (or other relatives) an animal has, the
higher the accuracy tends to be. Conversely, a producer interested in
a young boar with an excellent EPD, but a lower accuracy, should be
aware that his offspring could be much better or worse than predicted
and could cause his EPD to go much higher or lower upon reevalua-
tion. Table 1 contains ranges of possible changes in EPDs, depending
on accuracy, for four traits evaluated by the STAGES program for
Yorkshires. The publication NSIF - FS8 has more details on how
accuracy values are determined.


Table 1. Possible changes (+/-) associated with accuracies for
         maternal and growth traits evaluated by the STAGES Program
         for the Yorkshire breed.

                            Possible Change
         Accuracy    NBA     LW21    DAYS    BACKFAT
           .10       0.52    9.51    5.80    0.06
           .20       0.46    8.45    5.16    0.05
           .30       0.40    7.40    4.51    0.04
           .40       0.34    6.34    3.87    0.04
           .50       0.29    5.28    3.22    0.03
           .60       0.23    4.23    2.58    0.03
           .70       0.17    3.17    1.93    0.02
           .80       0.11    2.11    1.29    0.01
           .90       0.06    1.06    0.64    0.01

         To use the table, find the accuracy reported for an animal
         (column 1). Subtract the appropriate value for the trait
         being evaluated (columns 2-5) from the reported EPD to get
         the lower end of the possible change, and add the same amount
         to get the upper end of the possible change. A new EPD on the
         animal should fall within that range under normal

         Example: Two sires have the same EPD of +.25 pigs born alive
         (NBA). Sire A has an accuracy of .30 and Sire B has an
         accuracy of .80. Thus the possible change for Sire A is .25 +
         .40 while the possible change for Sire B is .25 + .09.


Interpreting Across-Herd Evaluations

      Across-herd evaluation is simply a tool that will help in making
selection decisions. Therefore it is critical that a producer know how
to get the best use of this tool.

      Know the genetic base. Currently, EPDs are based on breed
averages.  Thus animals in different breeds cannot be compared based
on EPDs. The genetic base may also change over time, so EPDs published
at different times cannot be compared. And EPDs generated by different
programs (for example, STAGES and the central test sta- tions) cannot
be compared because different animals are included in each evaluation.
In fact, the best way to make comparisons is to compare the magnitude
of differences in EPDs among animals on the same list. This is
particularly important for seedstock producers who need to compare
their own animals to those in other herds.

      Balance multiple traits. If more than one trait is of interest,
the importance of each trait will have to be weighted, because in most
cases animals will have better EPDs for some traits than others;
rarely will an animal be outstanding in all traits. Some programs
similar to STAGES offer indexes which combine EPDs of several traits
based on the relative accuracy of the estimates and an assigned
relative economic importance.  Individual producers must decide,
however, whether the indexes fit their particular situations. Other
programs report only EPDs for each trait, leaving it up to each
producer to decide which trait(s) should carry the most weight. Such
decisions include the same factors (accuracy and economic value) used
in the indexes, but indivi- duals must make the weighting decisions.

      Know how traits are reported. Negative EPDs can be favorable for
traits like days to market, whereas traits like average daily gain
should be positive. In this case, both evaluate growth potential, and
breed average is zero.

Application of Across-Herd Evaluations

      In addition to knowing how to interpret EPDs from across-herd
evaluations, it is important to realize that EPDs may play different
roles in different herds. For example, purebred breeders and
commercial operators will probably use EPDs differently.
Implementation of across- herd evaluations may also impact on the
structure of the seedstock industry.

      Seedstock producers. For the most timely selection decisions,
across- herd genetic evaluations should be used in conjunction with
within-herd evaluations.  The across-herd evaluations will enhance
within-herd performance testing programs, which may include STAGES
within-herd analyses, SPIs, NSIF selection indexes, or other methods
which rank animals on genetic merit relative to the average of their
contemporary groups (see NSIF- FS8, Estimating Genetic Merit).

      Within-herd evaluations normally have faster turnaround times,
but are less accurate, than across-herd ana- lyses. Thus a producer
may wish to use within-herd information to make preliminary selection
decisions, such as which gilts to retain at market weight. The
producers may use across-herd evaluations to fine tune the decisions,
such as whether to keep bred gilts or sell them.

      Another possibility would be to use across-herd evaluations to
compare potential breeding stock within one herd to pigs in other
herds, or in central test stations (see NSIF - FS11, Utilizing Central
Boar Test Sta- tions to Enhance Genetic Progress). Across-herd
comparisons allow breeders to identify herds and lines that are
genetically superior for the traits in which they are interested. A
breeder can locate the best sires and dams in the breed, then use
within-herd analyses to evaluate sons and daughters that are currently
available.  A breeder can decide whether to find replacements from
within the herd, or to purchase animals from a dif- ferent source. In
other cases, producers may be interested in evaluating traits not
included in the across- herd evaluations; within-herd evaluation is
the only possibility.

      Across-herd EPDs provide seedstock producers with a better idea
of strengths of their own animals on which they should build, relative
to other herds within a breed. In addition, EPDs provide the most
accurate infor- mation possible for making culling decisions, deciding
on how boars should be used in the breeding program, and determining
the value of animals offered for sale. EPDs alert seedstock producers
to areas needing improvements.

      Commercial producers. Across-herd evaluations can be used to
improve commercial operations, albeit in a dif- ferent manner.
Although crossbreeding is economically critical for commercial
operations, it does not result in on-going genetic change. Continual
genetic progress in commercial herds is directly dependent on progress
made in the operations from which breeding animals are purchased.
Commercial producers should choose seedstock sources naming similar
goals and management, then purchase animals with the best EPDs
possible, given budget constraints. However, decisions must be made
within breeds, because EPDs on animals in one breed cannot be compared
to EPDs on animals in another breed.

      EPDs should not be the basis for deciding which breeds to use in
a crossbreeding program. Breed choices should be made based on breed
averages, maternal effects, and heterosis. Because they are based on
the average, half of the animals in each breed will have favorable
EPDs and the other half will be below average. However, the number of
published EPDs can help determine the availability of breeding stock
for a particular breed: the more animals evaluated, the more likely a
producer can find what is needed. Genetic Improvement of Sire and Dam
Lines for Enhanced Performance of Terminal Crossbreeding Systems (NSIF
- FS14) provides a discussion of combining the advantages of
crossbreeding with genetic selection.

      Table 2. EPDs (and accuracies) of three Yorkshire
               sires evaluated in 1991 *
      Sire            NBA     LWT     DAYS    BF
      Ulf             .54     4.97    -.53    - .11
                      (.60)   (.62)   (.62)   (.74)

      California      .21     4.69    -1.39   - .05
                      (.59)   (.60)   (.56)   (.70)

      K. D. Boss      .51     5.51    .54     - .02
                      (.62)   (.63)   (.53)   (.58)
      * Adapted from the July 1991 American Yorkshire
        Club STAGES National Sire Evaluation Trait
        Leaders list.

      Once the choice of breed is made, then individual animals within
each breed should be chosen based on EPDs. A commercial producer can
use the across-herd analyses to locate superior sire and/or dam lines,
then use the within-herd analyses to select among the available sons
and daughters. For example, Table 2 contains EPDs on number born alive
(NBA), litter weight (LWT), growth (DAYS) and backfat (BF) for three
Yorkshire boars evaluated in 1991. All three have similar accuracies
for each trait, so decisions can be made based on EPD. If a producer
is interested in improving one trait, say NBA, then EPDs for NBA would
be most important in making decisions. In this case, Ulf and K. O.
Boss have similar EPDs for NBA, and are better than California. How-
ever, it is always wise to note EPDs on other traits, because
improvement in one should not result in a loss in another. For both
boars, EPDs for LWT (positive) and BF (negative) are favorable, but
DAYS is a different story: favorable (- .53) for Ulf, but unfavorable
(+.54) for K.D. Boss. Thus, a producer would have to deter- mine how
important growth is when deciding whether to purchase sons or
daughters of these two boars. On the other hand, if growth is very
important, a producer may want to consider progeny of California. Even
though the EPD for NBA is less than those of the other two boars, it
is favorable, as are the EPDs for LWT and BF.  The indexes provided by
STAGES can sometimes be helpful in making such decisions, but
producers must under- stand how the indexes are formulated. The
Yorkshire breed association has material which provides such infor-
mation, and producers need to be familiar with it before using the

      Industry structure. Across-herd evaluations will enable breed
associations to implement true national breed tests that could
incorporate the current central testing organization with on-farm
evaluations of additional animals. This would allow herds from across
the country to merchandize animals without physically transporting
them to central locations, because EPDs are comparable. It should be
noted, however, that EPDs will not replace a critical assessment of
physical and reproductive soundness, breed type, and conformation.

      Across-herd EPDs would also allow more accurate comparisons of
small groups of pigs, especially in central test situations in which
several breeds may be on test concurrently. If a test station has only
one pen of a breed, for instance, the only way to index the animals is
to use the average of all pigs of all breeds in the station at the
same time. Because all breeds are not equal in the growth and carcass
traits, however, that one pen may be severely penalized when ranked on

      For larger operations concerned about the risk of disease,
across-herd evaluations will permit the evaluation of all breeding
stock in each of several nucleus herds, and could allow for the
incorporation of data from multiplier herds to increase the accuracy
of evaluation. Cooperative arrangements among independent breeders
with herds made up of animals of similar genetic backgrounds could
work in the same manner. Such agreements would allow the producers to
take advantage of having large numbers of performance tested animals
to merchan- dise, without relinquishing individual ownership.

      To fully realize the potential of across-herd evaluations, a
viable AI industry is vital, whether it be independent boar studs, AI
centers, or a combination thereof. Embryo transfer and exchange of
breeding stock among seedstock producers will also contribute to the
success of across-herd evaluation programs. In all cases, accurate
identification of semen, embryos and animals is essential, as well as
complete performance records on resulting offspring. The NSIF
Guidelines contains information on standardization of records, the use
of which would immensely improve the efficiency of genetic


      Across-herd genetic evaluations could significantly increase the
rate of genetic progress in traits of economic importance to the swine
industry.  Several programs are available to seedstock producers. The
pro- grams will improve as more animals are added to the data base. To
fully realize the potential of across-herd evaluations, however,
seedstock producers, breed associations, AI companies, and others
interested in genetic improvement must work together to obtain
complete, accurate performance data for use in the programs.

NEW 2/92 (3M)


Cooperative Extension Work in Agriculture and Home Economics, State of
Indiana,  Purdue  University  and  U.S. Department  of  Agriculture
Cooperating. H.A. Wadsworth, Director, West Lafayette, IN. Issued in
furtherance of the Acts of May 8 and June 30, 1914. It is the policy
of the Cooperative Extension Service of Purdue University that all
persons shall have equal opportunity and access to our programs and