MAXBULL: COMPUTERIZED SIRE SELECTION FOR THE HERD

DAIRY HERD IMPROVEMENT
1986

CASSELL, B.G.
VOLUME: NCDHIP HANDBOOK

The development of artificial insemination (AI) and the
availability of frozen semen made possible the use of many sires in
a dairy herd. These agricultural
advances, although increasing flexibility, were both a blessing and
a curse. More bulls from which to choose meant more
decisions--decisions many dairy
producers felt unprepared to make. The number of bulls from which
to choose was not the only problem. Developments in sire evaluation
procedures created many
proofs for different traits; proofs which changed twice annually.
These details led some producers to believe that there was a
perfect mate for each cow and
breeding programs were developed to achieve the perfect matings.
Unfortunately, this approach often emphasized a cow's weaknesses
and deemphasized the
important herd breeding objectives.

Sire selection is critical to genetic change and is best
accomplished by planning for the needs of the entire herd. The
MAXBULL computerized sire selection program, developed at Virginia
Polytechnic Institute, assists dairy producers in choosing service
sires for their herds. MAXBULL, designed to store sire information,
maximizes sire merit for one of four traits: Predicted Difference
(PD) milk, PD dollars (PD$), and either of these traits minus its
confidence interval. Confidence interval is a statistical tool that
indicates the possible range of errors in estimating the PD's of a
bull. MAXBULL uses linear programming to ensure the selection of
the best group of bulls for specific breeding objectives. This
report summarizes the features of MAXBULL and provides background
information on the theoretical basis of the program.

Methods of Sire Selection

Sire selection follows one of two general approaches. The first,
truncation selection, is the simplest, but also the least
efficient. Truncation selection requires that each bull used in a
breeding program exceed specified minimums for each of the several
traits considered by a breeder. To alter the importance of traits,
the breeder must adjust the lowest acceptable level of each trait.
The major weakness of this approach is that there is no
consideration given to the bull that is outstanding for all but one
trait. Real inefficiency results when the bull is barely below
standards for that one trait. Survivors of truncation selection
often are bulls without strengths or weaknesses.

Index selection is the second approach to choosing sires. Under
this approach, different weights are applied to each trait
considered important to dairy producers. Variation in traits is
equalized through the application of weights. The numbers
representing the weighted traits are added to produce a single
number. This number is used to rank the bulls, and the higher
ranking bulls are selected as sires. This approach allows strength
in an important trait to offset weakness in some other trait and is
more efficient than truncation selection. Index selection is
effective, but it has some problems in practical situations. For
instance, the weights often are difficult to establish and may vary
from herd to herd. Also, index selection creates operational
difficulties because one bull will be best for all dairy producers
that have the same priorities.

The MAXBULL computer program solves the problem of establishing
correct economic weights and of determining how much to use each
individual bull. With this approach, each dairy producer
establishes herd breeding goals for fat test, overall type, semen
price and several individual type traits. The user then selects one
of four economically important traits to receive primary emphasis
in the selection program. (These four traits will be explained
later.) The computer chooses bulls with the specified minimum
averages for the fat test and physical type traits and with the
specified maximum average semen cost. Most important, the computer
determines these averages for bulls with the highest average for
the primary economic trait chosen by the producer. The result is a
recommended inventory of semen. This inventory suggests which bulls
to use and how much to use each bull in order to achieve the herd
breeding goals.

The MAXBULL approach can be superior to either the truncation or
index selection methods. Through the use of the MAXBULL computer
program, each dairy producer sets individual herd breeding goals.
MAXBULL allows an individual bull's strengths to offset his
weaknesses, and it allows strengths in one bull to offset
weaknesses in another bull. The program also recommends the rate of
usage for each bull. Linear programming techniques are used to
obtain an optimum solution. Optimum means that the combination of
bulls selected have the highest possible genetic merit among active
AI bulls for the trait considered by the dairy producer to be of
primary importance. The bulls chosen are optimum for the herd
breeding goals of test, physical type traits and price.

Selecting Traits and Establishing Goals

The trait to optimize is chosen from one of the following; PD milk,
PD$ and each of these minus its confidence interval. Although
confidence intervals can be calculated for every trait, they are
printed only for PD milk and PD$ by MAXBULL. Confidence intervals
become smaller as a bull's Repeatability (the reliability of the
PD's) increases. This means smaller errors are expected to be made
in estimating the breeding value of a bull as more and more
daughter information is available. Subtracting the confidence
interval produces a value that the true breeding value of the bull
will exceed 80 percent of the time.  Bulls with lower
Repeatabilities have larger confidence intervals. Subtracting
confidence intervals before making selection decisions means that
dairy producers will use low Repeatability bulls only when those
bulls' evaluations are high enough to justify the risk. For most
producers, PD$ is probably the best trait within their herds to
maximize. If changes in a bull's proof can cause problems for
producers selling breeding stock, PD$ minus its confidence interval
may be a better choice. Some producers may have personal
preferences for maximizing the PD milk.

Once the trait to maximize is selected, the important job of
establishing herd breeding objectives begins. It is important not
to place excessive emphasis on fat test or physical type traits or
both test and type. This excessive emphasis reduces the average
sire merit for the trait being maximized. Dairy farmers should
consider the following guidelines in establishing herd breeding
goals.

Fat percentage. Most herd breeding goals are around 0.0 for this
trait. If PD$ is maximized, the fat percentage takes care of
itself, but this goal is important if PD milk is being maximized.
The higher the herd average for milk production, the greater the
return on increased fat percentage because payment is made on
volume of milk. Higher goals for fat percentage should be used if
herd average is 2,000 pounds above breed average, if the Cow Index
(CI) for fat percentage is 0.02 or more below the average of breed,
or if the farmer relies on sale or breeding stock for substantial
income. Lower goals for fat percentage should be used if the
opposite is true.

Type. Setting herd breeding goals for type is subjective because
the value of type to commercial milk production is difficult to
quantitate. If breeding stock is sold from a farm, selection for
improved type increases returns. Breed averages for PD type (PDT)
in Holsteins is around 0.6. If no breeding stock is sold, goals of
0.0 to 0.6 are proper. Sale of 10 percent of replacement heifers
for breeding purposes would justify goals of 0.4 to 1.0. If all
heifers are sold, as in a heifer rearing enterprise, goals for PD
type could be from 1.0 to 1.5. If breed average differs from 0.6,
adjust these recommendations accordingly.

Semen Price. MAXBULL uses maximum semen price as a goal. The proper
goal for a herd depends on conception rate, heifer management
practices and sale of breeding stock. Herds with better conception,
lower calf mortality and younger ages at heifer freshening can
afford to pay more for semen. Also, herds selling more breeding
stock can afford to pay more for semen. Price goals for commercial
herds vary from $10 to $15 per unit of semen if conception rates
are approximately 50 percent and calf mortality is 5 to 15 percent.
If a herd has higher conception rates, a higher price can be paid
for semen. If approximately 10 percent of its heifers are sold from
a herd for breeding purposes, the per unit price of semen can be
increased by $5 to $10.

Linear type traits. MAXBULL allows herd breeding goals for udder
support, rear udder, fore udder, teat placement, rear legs (side
view) foot angle and stature. Herd breeding goals are the
"percentage of semen from improvers". MAXBULL provides conservative
estimates. Each unit of semen from an inferior bull is offset by a
unit of semen from a superior sire before the specified surplus of
semen from improvers is considered. Bulls with traits registered in
the top 20 percent of their breed are considered improvers. Bulls
with traits that fall in the bottom 20 percent are considered
inferior. For hind legs, bulls in the middle range are coded as
improvers while bulls at either end of the spectrum are considered
inferior. Herd breeding goals for linear type traits should be kept
to a maximum of 10 to 15 percent because the financial rewards of
such selection practices may not be realized. If several traits are
listed as needing substantial improvement, average sire merit for
the trait maximized can plummet!

Setting herd breeding goals in order to use MAXBULL may be a new
experience for most dairy producers. Producers using MAXBULL should
not try to develop perfect herd breeding goals on the first
attempt. Goals can be modified and rerun after initial results are
obtained. After satisfactory goals are developed, these goals will
be stored in permanent computer files and automatically rerun each
time new bull proofs are published. The use of herd breeding goals
is a fairly new concept in sire selection, and experience is
necessary to tailor these goals to the needs of a herd. Rerunning
MAXBULL until herd breeding goals are acceptable can take some
effort, but is critical to the success of the program. The quality
of the bulls recommended by the computer is determined by the goals
set by the dairy farmer. The computer does not think, it calculates
and organizes data. To be effective, herd breeding goals must be
continually reevaluated. Because the generation interval for dairy
cattle is at least 4 years, genetic progress in the dairy herd is
slow. Time spent formulating herd breeding goals is a wise
investment in herd improvement.

Current Use of MAXBULL

The MAXBULL sire selection program is run routinely for about 350
dairy producers in Virginia and other states. In January 1985,
there were 405 Holstein bulls available for use in the MAXBULL
program. These sires averaged $75 for PD$, 0.00 for PD fat
percentage, and.66 for PD type. Average semen price was $11. When
the MAXBULL program was run for the 315 Holstein herd breeding
goals on file, the average genetic merit of bulls recommended was
over $130 for PD$, above 0.01 for PD fat percentage, and 0.90 for
PD type. The average price per unit of semen recommended was $16.
Averages for test, type, and price reflect the desires of dairy
producers. The average sire merit for PD$ reflects the computer's
ability to organize the information in an optimum fashion. Higher
averages for PD$ were achieved for many individual farms as herd
breeding goals for test, type, individual type traits, and semen
price varied.

Access to the Program. The MAXBULL program is restricted to
mainframe computer use because memory requirements are large and
the bull file changes twice yearly. Currently, MAXBULL results are
available through the Dairy Science Department at Virginia
Polytechnic Institute where it was developed, first implemented and
continues to be improved. For more information, contact: Dr. Bennet
G. Cassell, Department of Dairy Science, Virginia Polytechnic
Institute, Blacksburg, VA 24061.

Herd breeding goals are maintained on computer files. When new sire
evaluations are published twice each year, MAXBULL files are
updated, new semen inventories are developed and mailed to dairy
farmers. The MAXBULL program offers a sound approach to genetic
improvement. The program ensures that all bulls available to a
dairy farmer are considered for use by: 1) evaluating the most
recent data, 2) controlling semen costs, 3) maintaining consistent
sire selection policies over time, and 4) optimizing sire merit for
the dominant trait. The program removes much of the burden of
acquiring new sire evaluations and evaluating all such proofs
simultaneously. MAXBULL can make managing sire selection easier as
well as successful.

Check with your state DHIA manager or Extension specialist for the
availability of MAXBULL in your state.

Other Features of MAXBULL

The previously mentioned features of MAXBULL do not allow
sufficient control of sire selection to produce manageable semen
inventories under all conditions. Consequently, additional features
have been added to the program. These features produce inventories
more useful to individual dairy farmers under specific conditions.
It is not critical that each user fully understand all of these
options. Presented for completeness and to demonstrate the
program's range of possible applications, a brief description of
these features follows.

Breed. MAXBULL can be run for Ayrshire, Guernsey, Holstein, Jersey
and Brown Swiss sires. The advantages of the program are greatest
for the breeds with the larger number of bulls.

AI organizations to use. Not all dairy producers want to buy semen
from all A.I. units. Each MAXBULL user specifies which AI
organization to use in the semen inventory. Bulls from all other
organizations are immediately eliminated from consideration.

Minimum semen from specific AI organizations. Some dairy producers
wish to buy semen from more than one AI organization, but wish to
buy most of their semen from one or two organizations. MAXBULL
allows each user to specify a minimum percentage of semen inventory
from one or two AI units. Caution: This feature can be too
restrictive if abused. Telling the computer to obtain 60 percent of
semen from one AI organization and 40 percent of semen from another
is mathematically restrictive if not impossible to solve. Users
should leave the computer some freedom by specifying 50 percent
from one AI unit and 30 percent from the other or 60 percent from
one AI unit only.

Semen price discounts. Prices used in the MAXBULL program are
direct herd semen prices for dairy producers in the southeastern
United States. They are based on prices from the most recent Sire
Summary. These prices may not apply to all herds, particularly
those participating in young sire sampling programs. Each user can
specify a price discount applicable to all prices for bulls in one
or two AI organizations so that the computer will select bulls
using prices more appropriate for a particular herd. This feature
also helps account for any patronage dividends.

Control of inventory. MAXBULL allows each user to specify how much
semen to include in an inventory.  The user also specifies the
minimum number of bulls to include. MAXBULL almost always provides
more bulls than the minimum, so this number should be kept low.
Either four or five bulls for moderately sized herds are
sufficient. Finally, MAXBULL allows the user to add to the usage of
a bull in multiples of semen. For instance, adding multiples of 5
(a goblet of semen) means the solution would recommend 5, 10, 15 or
so units of semen from each bull. This feature restricts the
computer, however. An inventory of 100 units filled 5 units at a
time means 20 separate decisions rather than 100 decisions.

Control of risk through Repeatability. Risk can be controlled three
ways in the MAXBULL program. The first is by maximizing a trait
minus its confidence interval as discussed earlier. The second is
by eliminating bulls below a designated Repeatability level to be
discussed in the bull restrictions section. The third way is by
limiting the use of all or some low Repeatability bulls. MAXBULL
allows the user to define a low Repeatability bull. A value of 50
percent is the default. Each user can then specify the maximum
percentage of semen in the inventory to come from all bulls below
that level and the maximum percentage of semen to come from any one
such bull. An example would be to allow up to 25 percent of
inventory to come from bulls below 50 percent Repeatability, but to
allow no more than 10 percent of inventory to come from any one
such bull.

Bull restrictions. MAXBULL users can establish a series of edits to
eliminate bulls below specified standards for Repeatability, PD
milk, PD fat percentage, PD type, PD$, and an upper limit for price
per unit of semen. These edits can serve a useful function,
particularly for maximum price per unit of semen. However, because
the edits are applied simultaneously for all traits, it is easy to
eliminate almost all bulls from consideration. Users should use
very low edits like 200 pounds for PD milk, -.30 for PD fat
percentage, $50 for PD$, and -1.0 for PD type. The computer will
select bulls meeting specified averages for test, type, and price
while maximizing PD$ or PD milk. Thus, edits for these traits can
be ignored without jeopardizing sire merit.

Restrictions on individual bulls. MAXBULL may allow no more than
four bulls to a specified percentage of semen inventory. This
feature can be used to restrict the use of bulls which are carriers
of undesirable recessive genes, bulls used heavily in previous
months, or bulls in limited semen supply.

Restrictions on sons of specific sires. The AI sire population
consists of families with similar sires. In some cases, MAXBULL
users may wish to restrict the use of semen from sons of certain
bulls to avoid inbreeding problems. The user of MAXBULL may specify
up to four sires to be restricted. A user also can use this feature
to identify sons of certain sires without limiting their use.
a



תתתתתתתתתתתתתתתתתתתתThe National Dairy Database (1992)תתתתתתתתתתתתתתתתתתתת
תתתתתתתתתתתתתתתתתתתתתתת\NDB\DAIRY\TEXT\DA103700.TXTתתתתתתתתתתתתתתתתתתתתתתת


%f TITLE;MAXBULL: COMPUTERIZED SIRE SELECTION FOR THE HERD
%f COLLECTION;DAIRY HERD IMPROVEMENT
%f ORIGIN;Virginia
%f DATE_INCLUDED;June 1992