usda logo

UNPUBLISHED
RESEARCH REPORT
1994


POSTED
AUGUST 2004

Age-Season Standardization for Yield Traits

M. M. Schutz*
Animal Improvement Programs Laboratory, ARS-USDA, Beltsville, MD 20705-2350
301-504-8334 (voice) ~ 301-504-8092 (fax) ~ rlaipl@aipl.arsusda.gov ~ http://aipl.arsusda.gov

*Current address: Department of Animal Sciences, Purdue University, West Lafayette, IN 47907 ~ 765-494-9478 (voice) ~ 765-494-9347 (fax) ~ mschutz@purdue.edu

Financial support from National Association of Animal Breeders with contributions from American Jersey Cattle Association, National Dairy Herd Improvement Association, Holstein Association USA, American Guernsey Association, and Brown Swiss Association is gratefully acknowledged.

Former adjustment factors for calving age and season are nearly 20 years old. Those factors were published in 1974 for milk and fat and in 1979 for protein for Holsteins. Fat factors were used to adjust protein records for other breeds. In the past 20 years, management improvements have resulted in less seasonal calving, earlier maturity of cows, and especially in reduced impact of summer heat and humidity. Also, former factors were developed from models that did not consider relationships among animals or differences due to genetic trend. Within a given time period, younger cows are genetically superior to their older herdmates. If genetic relationships are not considered, some of this genetic superiority is reflected in age solutions, which thereby are overestimated. Thus, resulting standardization factors would be too small. Research showed that former factors overadjust records of young cows, but the overestimation is apparently less when genetic relationships are considered than was thought based on cow repeatability models.

Current research at the Animal Improvement Programs Laboratory (AIPL) has been directed toward estimating adjustment factors for milk, fat, and protein with animal models and relationships. Other objectives of this work are to account for differences in age effects over time; to determine whether adjustment should be for differences in record means (additive), variances (multiplicative), or both; and to determine how factors are affected by parity and calving interval.

Beginning January 1995, parity (for the first 5 lactations) and age (3 groups) within parity will be included in animal models used for genetic evaluation. Recent work at AIPL and in France has shown that inclusion of parity in animal models reduced estimates of genetic trend that were too high (as high as phenotypic trend despite improvements in feeding and management) and that parity should be included in animal models even after new age standardization factors are employed.

Inclusion of parity and age within parity in animal models will enhance genetic evaluations but will not eliminate the need for accurate standardization factors. Additive factors to standardize records adjust means but not variances. Multiplicative factor standardize both means and variances. Proposed inclusion of parity and age within parity directly in animal models is an additive approach. However, some degree of multiplicative adjustment appears necessary to standardize milk, fat, and protein records because there is a mean-variance relationship. As cows mature and mean production increases, the variance of production also increases. Preadjustment by multiplicative factors, coupled with the additive approach of including parity and age within parity in animal models for genetic evaluation, appears to be appropriate.

Former age standardization factors did not consider the effect of lactation number. For example, a second-lactation cow calving at 44 months got the same adjustment as a third-lactation cow calving at 44 months. Comparisons of first- and second-parity cows calving at the same age were also wrong, but the problem was less pronounced for genetic evaluations because our animal model normally includes these in separate management groups. However, if there are too few management group mates, groups may be combined over lactations. Proposed parity and age-within-parity effects are intended to remove parity differences, especially for second or later lactations, and to account for most differences in age effects from year to year and for changes arising during time periods between estimation of age adjustment factors.

Season effects are also important for milk production. In animal models, effects of season primarily are accounted for by management groups. However, this is not very effective when group mates are few and management groups are combined across several months. In this case, some preadjustment for season of calving is necessary. Over time, production loss associated with summer calving has become less severe, especially in the Southeast. Age-season interaction is also known to be important but is not considered directly in animal model evaluations. Seasonal influences are different for cows of different parities. These effects are largest in regions with hot humid summers. Preliminary work suggested that age-season interaction has changed over time.

Changes in Age-Season Standardization for Genetic Evaluations

Parity

To preadjust for parity and age within parity effects, factors for age at calving will be separate for first, second, third, fourth, fifth, and later lactations. Only adjustments for the first 5 parities are needed to adjust records for genetic evaluations. Separate factors will continue to be used for each calendar month. Former factors were developed from the combined effects of age classes across lactations and age group-month interaction, where age groups did not consider lactation number but served as broader categories of ages to reduce the number of age-month subclasses required. New factors will be for the combined effects of age classes within parity and parity-month. Thus, age-season interaction will be a parity-month effect rather than an age group-month effect as was previously considered.

Animal Models

New age-season standardization factors are from animal models that make use of known relationships and account for differences in genetic trend. The following model was used to obtain solutions for age-season effects:

y = TRLM + TRC + TDc + TDp + HY + P + A + e

where: y = milk, fat, or protein record
  TRLM = time, region, parity, month of calving
  TRC = time, region, age class within parity
  TDc = time, class for days open in current parity
  TDp = time, class for days open in previous parity
  HY = herd, year
  P = permanent environment (random)
  A = animal (random)
  e = residual (random)

Age-season solutions are formed by combining effects of parity-month and age class within parity for each time and region. Solutions were obtained from single trait JAA program of Dr. Ignacy Misztal. For each trait, all regions and time periods were included concurrently for breeds other than Holstein. Because of computational constraints, Holsteins were split into 7 sets (regions 1 and 2, regions 3 and 6, regions 4 and 5, regions 7 and 9, region 8, regions 10 and 11, and region 12). Geographic location and number of records were considered when combining regions.

Time periods

Age-season factors have changed over time. Separate factors were estimated for 5 time periods: 1964-68, 1969-74, 1975-80, 1981-86, and 1987-92. The first time period is the one from which former factors were derived. The last year in which all records had the opportunity to reach 305 days was 1992. For protein, the number of records to estimate adjustment factors adequately was available only for the last 2 time periods. Only the final time period had enough Holstein records for Puerto Rico. Separate factors will be estimated for the 5 time periods. Factors from the most recent time period will be used to adjust records initiated after 1992, and records from the first available time period will be used for earlier records.

Regions

Factors are separate for geographic regions within breeds. For current factors, number of regions ranged from 1 (national) for Milking Shorthorns to 11 for Holsteins (12, if more recently developed factors for Puerto Rico are considered). Region-to-region differences will be less pronounced than for former factors. Region definitions will remain the same with 2 exceptions. Because of fewer Guernseys, there will be 4 instead of 6 regions, and regions for Guernseys will be the same as the current Ayrshire and Brown Swiss regions. Definition of Jersey regions also will change but number of regions will remain the same. Regions will be defined as follows:

Ayrshire, Brown Swiss, and Guernsey
Region 1 Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont
Region 2 Alabama, Arizona, Arkansas, California, Delaware, Florida, Georgia, Hawaii, Kentucky, Louisiana, Maryland, Mississippi, Nevada, New Mexico, North Carolina, Puerto Rico, South Carolina, Tennessee, Texas, Virginia, West Virginia
Region 3 Colorado, Illinois, Indiana, Iowa, Kansas, Michigan, Missouri, Nebraska, Ohio, Oklahoma
Region 4 Alaska, Idaho, Minnesota, Montana, North Dakota, Oregon, South Dakota, Utah, Wisconsin, Washington, Wyoming
Holstein
Region 1 Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont
Region 2 New York
Region 3 New Jersey, Pennsylvania
Region 4 Delaware, Kentucky, Maryland, Virginia, West Virginia
Region 5 Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas
Region 6 Indiana, Michigan, Ohio
Region 7 Colorado, Illinois, Iowa, Kansas, Missouri, Nebraska, Oklahoma
Region 8 Wisconsin
Region 9 Minnesota, North Dakota, South Dakota
Region 10 Arizona, California, Hawaii, Nevada, New Mexico
Region 11 Alaska, Idaho, Montana, Oregon, Utah, Washington, Wyoming
Region 12 Puerto Rico
Jersey
Region 1 Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont
Region 2 Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, Puerto Rico, South Carolina, Tennessee, Texas, Virginia, West Virginia
Region 3 Illinois, Indiana, Michigan, Missouri, Ohio
Region 4 Iowa, Minnesota, Nebraska, North Dakota, South Dakota, Wisconsin
Region 5 Arizona, California, Colorado, Hawaii, Kansas, Nevada, New Mexico
Region 6 Alaska, Idaho, Montana, Oregon, Utah, Washington, Wyoming
Milking Shorthorn
Region 1 National

Age Classes

Age classes are in the following table. Eight age-at-calving classes were assigned within lactations 1, 2, and 3. Obviously, fewer records are available with increasing lactation numbers. Therefore, 6 classes were assigned within lactation 4, and 5 classes were assigned in lactations 5 and 6 or later. Most classes within lactations contained more than a single age in months, with the exceptions being for 24, 25, and 26 months of age in lactation 1 and 36, 37, and 38 months of age in lactation 2.

Parity 1 Parity 2 Parity 3 Parity 4 Parity 5 Parity 6
Class Ages Class Ages Class Ages Class Ages Class Ages Class Ages
1 18-21 9 28-33 17 40-46 25 52-60 31 64-73 36 76-86
2 22-23 10 34-35 18 47-48 26 61-63 32 74-77 37 87-96
3-5 24-26 11-13 36-38 19 49-50 27 64-65 33 78-80 38 97-120
6 27-28 14 39-40 20 51-52 28 66-68 34 81-84 39 121-144
7 29-31 15 41-43 21 53-54 29 69-71 35 85-91 40 145-200
8 32-35 16 44-49 22 55-56 30 72-77    
    23 57-58  
24 59-63

Days open

Because effects of age at calving are influenced to a large extent by calving interval in previous lactations, it is very important to account for days open in determining age at calving solutions. For example, cows calving at an early age within a parity would often be disadvantaged by short days open in the preceding lactation. New age factors are from animal models that consider effects of days open in the previous and current lactation. Research has shown larger effects on age at calving from previous days open than from current days open. Traditionally, reporting of reproductive information (reported as days carried calf) has not been complete. When a subsequent calving date is known, days open can be verified or approximated, but terminal records may not have days open for the current lactation. Of course, previous days open must be treated as undefined for first lactations and unknown for initial lactations of cows entering test at later ages. Effects of days open were considered within time period for all breeds and within the 7 sets of regions for Holsteins. For current days open, 29 classes (20-29, 30-39, …, 290-299, and 300-305) were fit within first, second, and third or later lactations plus a single class in each lactation group for unknown days open. For previous days open, a single class was fit for all first lactation records, and second and third or later lactations had the same 29 classes as for current days open plus a single class for unknown days open. Previous days open effects derived from these models will serve as a basis to standardize production records for calving interval. Current days open effects should be used at some time in the future.

Calculation of factors

Age solutions within parity must be smoothed prior to calculation of multiplicative adjustment factors to eliminate inconsistencies that may arise from sampling variance. This was accomplished through linear and quadratic regression of age class solutions on average ages of the age classes within time, region, and parity. A month-of-calving effect was estimated for each time, region, and parity and included the intercept from regression on age. For standardization programs at AIPL, month effects, linear and quadratic coefficients for regression on age, previous days-open equations, and mean production for cows calving at the base age within time period and region are loaded directly into standardization programs. For each record, a month effect, coefficients for linear and quadratic regression on age, and a previous days-open equation is obtained based on year of calving, geographic region, and parity. The resulting adjustment is calculated as

yieldbase

yieldbase + MOC + AGE + PDO

where yieldbase is the average yield of cows calving at the base age for a time period and region, MOC is a month-of-calving effect, AGE is an age-at-calving effect, and PDO is a previous days-open effect. This is the factor that is multiplied by the 2X-305 or extended yield record to obtain standardized yield. This strategy eliminates the need for extensive use of lookup tables, because processing can be by breed and region so that only coefficients, equations, and means for a single breed and region are needed. These may be stored in memory to increase computing speed.

Base ages

Updating of age-season adjustment factors provides an excellent opportunity to reconsider the base age to which records are standardized. Only the United States, Australia, Canada, and Italy adjust records to mature age. Most countries adjust records to a first-lactation age of 24-30 months. Israel adjusts to 36 months or the age of average production. Interbull has recommended adjusting to average age. For the United States, mature production is now reached earlier than when former factors were developed. For Ayrshires, Brown Swiss, Guernseys, and Holsteins, maximum production is reached in fourth lactation at 72 to 77 months of age; for Jerseys, in fourth lactation at 61-63 months of age, and for Milking Shorthorns, in sixth lactation at 76-86 months of age. However, this later age may be more the result of sampling variance than a true breed difference.

Advantages and disadvantages of adjustment to other than mature age have been discussed previously (McDaniel, 1973). Apparently, the only real advantage to adjusting to mature age is that it is traditional. Adjusting to age of average production (37 months of age for Guernseys, Holsteins, and Jerseys or 42 months of age for Ayrshires, Brown Swiss, and Milking Shorthorns) has several advantages. Records adjusted to mature age are hypothetical as only a small percentage of cows reach mature age. On the other hand, adjustment to average age is more realistic because it puts records on the scale of an average cow in the herds. Also, adjusted and actual yields would be similar, on average, for a herd. With adjustment to average age, bull predicted transmitting abilities would better represent the actual superiority or inferiority of an average daughter over all her lactations rather than the hypothetical superiority or inferiority of mature daughters. Also, most records would be changed less by adjustment to average age than to mature age, which implies introduction of less bias when global factors are not exact for a particular herd, state, or year.

New adjustment factors will be to a single base age within breed for all regions and seasons. As recommended by the Council on Dairy Cattle Breeding (September 19-20, 1994; St. Louis, MO), AIPL will not change adjustment from mature age to average age now "to allow the industry time to consider the nature and effect of such a change." The Genetic Advancement Committee of Holstein Association USA recommended that AIPL give further consideration to adjusting milk, fat, and protein records to average age some time in the near future.

Examples

For former factors, largest seasonal differences existed in region 5 for Holsteins. To illustrate how new adjustment factors have changed, consider a Holstein heifer calving since 1987 in August in region 5 at 20 months of age and with an actual 2X-305 record of 20,000 pounds. The former adjustment factor for region 5/20 months/August is 1.44. Her adjusted record was 20,000 × 1.44 = 28,800. The new adjustment factor for time period 5/region 5/parity 1/20 months/August is 1.389. With new factors, her adjusted record would be 20,000 × 1.389 = 27,780 or 1020 pounds less than with former factors.

Now consider the previous heifer's counterpart calving at the same age in the same region in the same year and producing the same amount of actual milk but calving in January not August. With former factors (factor for region 5/20 months/January is 1.33), her adjusted production was 20,000 × 1.33 = 26,600. With new factors (factor for time period 5/region 5/parity 1/20 months/January is 1.350), her adjusted production would be 20,000 × 1.350 = 27,000. Notice, in January, adjusted production is actually 400 pounds higher with new factors. New factors for young ages of calving in favorable months tended to be slightly larger than former factors, largely because, as previously mentioned, new models better accounted for genetic trend.

To illustrate the impact of including parity differences in the adjustment procedure, consider 2 cows calving since 1987, in January in region 5 at 45 months of age with 120 days open in the previous lactation and with actual 2X-305 records of 20,000 pounds. Cow A is calving for the second time, but Cow B is calving for the third time. With former factors (factor for region 5/45 months/January is 1.05), adjusted production was 20,000 × 1.05 = 21,000 for both cows. With new factors, adjusted milk for cow A (factor for time period 5/region 5/parity 2/45 months/January is 1.350) will be 20,000 × 1.038 = 20,760. With new factors, adjusted milk for cow B (factor for time period 5/region 5/parity 3/45 months/January is 1.015) will be 20,000 × 1.015 = 20,300. Thus, the difference is 460 pounds. For cows calving at the same age, multiplicative adjustment factors were larger for first than second parity, slightly larger for second than third parity, smaller for third than fourth parity, and smaller for fourth than fifth parity.


References

McDaniel, B. T. 1973. Merits and problems of adjusting to other than mature age. Journal of Dairy Science 56:959-967.

Norman, H. D., T. R. Meinert, M. M. Schutz, and J. R. Wright. 1995. Age and seasonal effects on Holstein yield for four regions of the United States over time. Journal of Dairy Science 78:1855-1861.

Schutz, M. M., and H. D. Norman. 1994. Adjustment of Jersey milk, fat, and protein records across time for calving age and season. Journal of Dairy Science 72(Suppl. 1):267 (abstr. 1030).