J. Anim Sci. 2006. 84:2289-2298. doi:10.2527/jas.2005-715
© 2006 American Society of Animal Science
Test duration for growth, feed intake, and feed efficiency in beef cattle using the GrowSafe System1
Z. Wang*,2,
J. D. Nkrumah*,
C. Li*,
,
J. A. Basarab
,
L. A. Goonewardene
,
E. K. Okine*,
D. H. Crews, Jr. and
S. S. Moore*
* Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, Alberta, T6G 2P5 Canada;
and
Alberta Agriculture, Food and Rural Development, Lacombe Research Center, Lacombe, Alberta, T4L 1W1 Canada;
and
Alberta Agriculture, Food and Rural Development, 7000-113 Street, Edmonton, Alberta, T4H 5T6 Canada; and
and
Agriculture and Agri-Food Canada Research Centre, Lethbridge, Alberta, T1J 4B1
 |
Abstract
|
|---|
This study was conducted to determine the optimum test duration and the effect of missing data on accuracy of measuring feed efficiency and its 4 related traits ADG, DMI, feed conversion ratio, and residual feed intake in beef cattle using data from 456 steers with 5,397 weekly averaged feed intakes and BW repeated measurements taken over 91 d. Data were collected using the GrowSafe System at the University of Alberta Kinsella Research Station. The changes and relative changes in phenotypic residual variances and correlations (Pearson and Spearman) among data from shortened test durations (7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77, or 84 d) and a 91-d test were used to determine the optimum test duration for the 4 traits. The traits were fitted to a mixed model with repeated measures using SAS. Test durations for ADG, DMI, feed conversion ratio, and residual feed intake could be shortened to 63, 35, 42, and 63 d, respectively, without significantly reducing the accuracy of the tests when BW was measured weekly. The accuracy of the test was not compromised when up to 30% of the records were randomly removed after the first 35 d on test. These results have valuable and practical implications for performance and feed efficiency testing in beef cattle.
Key Words: beef cattle • feed efficiency • repeated measures analysis • test duration
|
INTRODUCTION
|
|---|
Feed costs represent approximately one-half of the total cost of production for most classes of livestock (Kennedy et al., 1993
), and it is the single largest expense in most commercial beef operations (Arthur et al., 2004). Improvement of feed efficiency should be a major consideration in most livestock breeding programs (Kennedy et al., 1993). Recently, residual feed intake (RFI) was recognized as a trait for measuring feed efficiency and was defined as the difference between an animal’s actual feed intake (FI) and its expected feed requirements for maintenance and production (Archer et al., 1997; Archer et al., 1999; Arthur et al., 2001). In practice, RFI is estimated based on an animal’s daily DMI, ADG, and BW.
Accurate measurements of FI, ADG, and BW on individual animals require a test over a period of time. Because management costs as well as feed costs increase as test length increases, it would be highly beneficial to the industry to identify an appropriate test duration to reduce the costs of measurement without compromising data accuracy and reliability. In North America, a 112-d test was considered an industry standard for testing bulls for rate of gain (Franklin et al., 1987; Kemp, 1990; Brown et al., 1991). Archer et al. (1997) and Archer and Bergh (2000) suggested that a 70- to 84-d test was adequate to get an accurate measure of RFI in sires of British breeds and other biological types. However, these recommendations were obtained without considering the nature of repeated measurements of FI and BW on the same subject over a period of time during a test.
The objectives of this study were to determine the optimum test duration for the measurements of average weekly ADG, DMI, feed conversion ratio [FCR; the inverse of the efficiency of gain (G:F)], and RFI using a mixed model with repeated measures analysis and to examine the impacts on data accuracy and reliability caused by missing observations.
|
MATERIALS AND METHODS
|
|---|
Animal, Diets, and Phenotypic Data
All animals in the study were cared for according to the guidelines of the Canadian Council on Animal Care (CCAC, 1993).
Four hundred fifty-six hybrid steers from an experimental population were used in this study. Steers were progeny of the University of Alberta Hybrid dam line produced over more than 10 yr and composed of crosses among 3 composite lines, namely Beef Synthetic 1, Beef Synthetic 2, and Dairy x Beef Synthetic (Berg et al., 1990). Beef Synthetic 1 was composed of approximately 33% Angus and Charolais, approximately 20% Galloway, and the remainder of other beef breeds. Beef Synthetic 2 was made up of approximately 60% Hereford and 40% other beef breeds. The Dairy x Beef Synthetic was composed of approximately 60% dairy cattle (Holstein, Brown Swiss, or Simmental) and approximately 40% of other breeds, mainly Angus and Charolais (Goonewardene et al., 2003).
Steers were born in spring of 2002, 2003, and 2004, and postweaning feedlot performance and feed efficiency tests were carried out over 3 yr from November 2002 to May 2005 at the University of Alberta Kinsella Research Station, using the GrowSafe automated feeding system (GrowSafe Systems Ltd., Airdrie, Alberta, Canada). In each year, steers were randomly assembled into 2 contemporary test groups (CGP) based on the observed capacity of the test facility. Contemporary test groups 1, 3, and 5 were tested earlier in each year followed by CGP 2, 4, and 6. Therefore, CGP 2, 4, and 6 were older and heavier than CGP 1, 3, and 5. The CGP includes the year and season effects combined. The average age and BW at the beginning of the experiment are given in Table 1.
The feedlot diets were slightly modified over the 3 years based on the availability of feeds (Table 2). Corn was used in yr 1, instead of barley and oats, because of a feed-barley shortage in that particular year. Measurements of daily FI and weekly BW data on individual steers were collected for a total of 77, 84, 84, 70, 91, and 91 d for CGP1, CGP2, CGP3, CGP4, CGP5, and CGP6, respectively. The detailed procedures for the feedlot test were described by Nkrumah et al. (2004). Briefly, 2 tests consisting of approximately 80 steers per test were conducted each year. Each animal was identified by means of a plastic tag located in the left ear. Before entry into the testing facility, each animal was fitted with a passive radio frequency transponder button (Allflex USA Inc., Dallas-Fort Worth, TX) encased in plastic ear tags at a position 5 to 6 cm from the base of the right ear with the button on the inside. The test facility was a shed with one long side open to provide access to 10 feeding bunks. All steers had been vaccinated for bovine viral diarrhea and clostridial diseases and treated with a pour-on parasiticide before entry into the test.
In yr 1, steers were fed free-choice a backgrounding diet of mainly alfalfa-brome hay with oats and supplemented with corn and feedlot mineral supplement to promote a growth rate of just under 1.0 kg/d for approximately 30 d. This period was followed by a 30-d pretest adjustment period, in which the amount of corn in the backgrounding diet was increased gradually to introduce the steers to the test diet and the feeding system. This was done to allow them to adapt to the diet and learn to feed from the test facility. Feed intake was measured for each animal using the GrowSafe automated feeding system, which has been validated and used previously (Basarab et al., 2003). Briefly, the system consisted of ten model 4,000 feed bunks, a data-logging reader panel connected to each feed node, a personal computer, and GrowSafe Data Acquisition and Analysis Software. Wireless communication (Model 4000 R/F) allowed for the transfer of data between the acquisition unit and a desktop computer in an office located approximately 100 m away. Daily feed intake for each animal as recorded by the GrowSafe system was determined using specially customized software. Data collected on occasions when the automatic monitoring system failed to function due to power failure, mechanical problems, or failure of a main computer board (1 to 2% per test) were excluded from all subsequent analyses.
Traits and Calculations
Four beef production efficiency traits (ADG, DMI, FCR, and RFI) were studied. The ADG and initial BW on test were estimated by a linear regression of the steer’s observed BW against days on test as
where bwit is the observed weekly BW of animal i measured on day t during the test, ait is the estimated initial BW of animal i on day t, bit is the estimated regression coefficient that is equal to the estimated ADG of animal i on day t, xt is the weekly BW measurement on day t, and eit is residual error associated with each observed body weekly weight bwit. The high R2 values (0.988 ± 0.011; 0.931 to 0.999) indicate that growth during this phase of the steer’s life was linear and that the choice of a linear regression model was appropriate. Midtest BW of each animal was estimated as
where mbwit is the midtest BW of animal i during the test period t in days, and ait and bit are as defined earlier.
Average daily FI of each animal was converted to daily DMI and then converted to ME in MJ/kg based on the DM and the MJ of ME/kg of DM content given in Table 2
. To make the results comparable with other research work, the FI in megajoules of ME per kilogram of DM was divided by 10 MJ of ME/kg of DM to give total DMI standardized to an energy density of 10 MJ of ME/kg of DM. The FCR was calculated as standardized daily DMI divided by ADG for each animal to give the kilograms of feed required for 1 kg of BW gain. The RFI for each animal was calculated as the difference between each animal’s actual standardized DMI and its predicted standardized DMI based on its metabolic BW (MBW0.75) and ADG according to procedures described by Arthur et al. (2001).
The ADG, DMI, FCR, and RFI for each animal were calculated weekly at 7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77, 84, and 91 d on test and used for further statistical analysis. The actual test length was 70, 84, 84, and 77 d for CGP1, CGP2, CGP3, and CGP4, respectively, and therefore, the data in these CGP were treated as missing observations from 70 to 91, 77 to 91, or 84 to 91 in the repeated measures analysis.
Statistical Analysis
In this experiment, the mean ADG and DMI of each animal were obtained for each week repeatedly over the test period. Data of this nature often have the following inherent characteristics: 1) measurements made on the same steer are more likely to be correlated than measurements taken on different steers, 2) 2 measurements taken closer in time on the same steer are likely to be more correlated than measurements taken further apart in time, and 3) variances of the repeated measures often change over time (Wolfinger, 1996; Littell et al., 1998; Templeman et al., 2002), and this is especially so in the case of BW gain (Wang and Goonewardene, 2004). The analysis of repeated measures data therefore requires appropriate accounting for correlations between the observations made on the same steer and heterogeneous variances among observations over time. In addition, due to the limitation of the test capacity, these steers were tested in 6 CGP over 3 yr. The differences for beginning test age, beginning test BW, ADG, and efficiency traits on the 70-d test period existed among these CGP (Table 1) and needed to be appropriately adjusted in the statistical analysis. Therefore, the 4 traits were analyzed using the MIXED procedure of SAS (SAS Inst. Inc., Cary, NC) as a repeated measures analysis, to allow for heterogeneous variances and correlations among different time intervals on test (Littell et al., 1998, 2000; Wang and Goonewardene, 2004). The model used for this analysis was as follows:
 |
where yijt is the ADG, DMI, FCR, or predicted RFI of steer i of contemporary group j at time t; µ is the fixed overall mean effect; ßj is the fixed CGP effect j (contemporary group 1 to 6); 
t is the fixed day t (7, 14, 21, 28, 35, 42, 49, 56, 64, 70, 77, 84, and 91 d) on test effect; (ß)jt is the fixed interaction effect of CGP j with day on test t; 
i(j) is the random effect of steer i within CGP j; eijt is a random residual error associated with yijt; and Vi is a block diagonal covariance matrix associated with steer i.
To allow for heterogeneous variances over time on test and correlations among them, the first order ante dependence [ANTE(1)] covariance model was chosen for this analysis based on Schwarz’s Bayesian information criterion. The covariance matrix of the best fitting model provided measures of the changes in variance over time during the test period. To make the variance changes comparable among the 4 traits across the test period, a relative change of variance, defined as the percentage difference between the variance obtained from the previous measurement and the current measurement divided by the variance obtained from the first measurement (7 d), was used as an additional criterion to assess the variance changes.
In addition to the changes and relative changes in the phenotypic residual variances over time, correlations (Pearson and Spearman Rank) among data from a shortened test (7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77, or 84 d) and a 91-d test were used as additional criteria to determine the optimum test duration. The Pearson and Spearman correlations were estimated using the CORR procedure of SAS. Finally, 6 datasets that were randomly missing 5, 10, 15, 20, 25, or 30% of FI observations after 35 d on test were generated and analyzed using the MIXED and CORR procedures of SAS, as described above, to examine the effects of the missing observations on the optimum test duration, compared with the full data set.
|
RESULTS
|
|---|
Test Duration for ADG
The changes and relative changes of phenotypic residual variance and Pearson and Spearman correlations of ADG observed at 7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77, 84, and 91 d are given in Table 3. The results show that the trend of variances reduction appears to fluctuate over the test period (Figure 1). The relative changes in variance reduction (Table 3) reduced to 1.94, 0.97, and 0.97% from 42 to 49, 49 to 56, and 56 to 63 d, then increased to 2.91, 2.91, and 1.94% from 63 to 70, 70 to 77, and 77 to 84 d, respectively, and again decreased to 0.97% from 84 to 91 d. The changes and relative changes of variances over the testing period did not provide a clear trend to determine appropriate test duration for ADG in this study. This result indicates that for ADG, a longer testing period and more measurements are needed to obtain an accurate determination of test duration. There were greater week-to-week variations in ADG at the beginning and end of the test period. The fluctuations in the week-to-week variances implied that the relative changes in variance were variable and could not be applied to a clear determination for an optimum test duration. Practically, because the interest is in how quickly the variances reduction stabilized and correlations approach unity, it is still possible to determine the optimum test duration for measuring ADG with the actual variances reduction and correlation coefficients.
View this table:
[in this window]
[in a new window]
|
Table 3. Changes and relative changes of phenotypic residual variances and Pearson and Spearman correlations for ADG, DMI, FCR, and RFI of shortened tests and the full test of 91 d1
|
|
View larger version (13K):
[in this window]
[in a new window]
|
Figure 1. Comparison of phenotypic residual variance changes for different scenarios of missing observations after 35 d for ADG. Full = no missing data; 10% = 10% data missing; 20% = 20% data missing; 30% = 30% data missing.
|
|
Thus, examining the Pearson and Spearman correlations (0.90 and 0.87, P < 0.01) between the 63 and 91 d test in Table 3, it appears that the test duration of ADG could be shortened to 63 d with very little loss in accuracy. Although this conclusion is somewhat unclear when considering the relative variance reduction results, the Spearman correlation (0.87, P < 0.01) should provide a reliable decision because it measures the correspondence between ranks (Steel et al., 1997) of the different test durations. In other words, the results indicated that adding extra BW measurements would not significantly reduce the amount of unexplained residual variation in ADG in the study. Therefore, a 63-d test length is recommended for ADG.
Test Duration for DMI
The changes and relative changes of phenotypic residual variances for DMI in Table 3 showed a dramatic initial reduction that stabilized after 35 d of the test (Figure 2). The relative changes of phenotypic residual variances becomes –0.48, 0.29, 0.82, –0.24, –0.09, 0.34, 0.29, and 0.96% for 35 to 42, 42 to 49, 49 to 56, 56 to 63, 63 to 77, 77 to 84, and 84 to 91 d, respectively. The trend of variance reduction (Figure 2) is clearer than the trend seen for ADG and showed that extending the duration of data collection beyond 35 d resulted in very little improvement in accuracy. Thus, 35 d of test should be sufficient to obtain a relatively accurate measure of DMI. Additional measurements for DMI beyond 35 d do not appreciably improve measurement accuracy. Pearson and Spearman correlations (Table 3) between shortened tests (28, 35, 42, 49, 56, 63, 70, 77, and 84 d) and 91-d test for DMI reached 0.90 (P < 0.01) at 28 d. Based on the correlations and variance changes, the shortened 35-d test duration for DMI can be recommended.
View larger version (15K):
[in this window]
[in a new window]
|
Figure 2. Comparison of phenotypic residual variance changes for different scenarios of missing observations after 35 d for DMI. Full = no missing data; 10% = 10% data missing; 20% = 20% data missing; 30% = 30% data missing.
|
|
Test Duration for Feed Conversion Ratio
Feed conversion ratio is a function of 2 component traits, DMI and ADG (defined as DMI/ADG). Therefore, the changes and relative changes of phenotypic residual variances are influenced by the trends of its 2 component traits. The variance changes for FCR are shown in Figure 3. Like the trend observed in Figure 1 for ADG, the reduction in phenotypic residual variance is very dramatic during the first of 4 wk (7 to 35 d) and becomes virtually flat after 35 d with mild fluctuations. The relative changes of phenotypic residual variances for FCR (Table 3) are less than 1% (0.45, 0.20, and 0.70%) from 42 to 63 d and greater than 1% thereafter (1.86, 1.56, and 1.41%) from 63 to 84 d. It becomes negative after 84 d (–1.05%). Based on the variance reduction, a 42-d test for FCR can be suggested. On the other hand, Pearson and Spearman correlations reached 0.90 (P < 0.01) between the 42- and 91-d test (Table 3). These results indicate that a 42-d test should be sufficient for FCR under the GrowSafe System when BW is measured weekly.
View larger version (15K):
[in this window]
[in a new window]
|
Figure 3. Comparison of phenotypic residual variance changes for different scenarios of missing observations after 35 d for feed conversion ratio. Full = no missing data; 10% = 10% data missing; 20% = 20% data missing; 30% = 30% data missing.
|
|
Test Duration for Residual Feed Intake
The trend in changes and relative changes of phenotypic residual variance (Figure 4) is similar to the trend in DMI (Figure 2). However, the phenotypic residual variances reduced dramatically in the first 3 wk followed by a gradual decrease thereafter. The relative changes in phenotypic residual variance become very small (0.74, 0.43, 0.80, and 0.49% for 63 to 70, 70 to 77, 7 to 84, and 84 to 91 d, respectively) after 63 d. This indicates that a 63-d test duration should be sufficient to obtain an accurate measure of RFI and therefore can be recommended. Increasing the number of measurements after 63 d does not provide more information or improve the accuracy of RFI. The Pearson (0.90, 0.95, 0.97, and 0.99) and Spearman (0.90, 0.95, 0.98, and 0.99) correlations among the shortened tests (63, 70, 77, and 84 d) and the 91-d test in Table 3 also support this finding. The correlations between a 63- and 91-d test for RFI both reached 0.90 (P < 0.01).
View larger version (16K):
[in this window]
[in a new window]
|
Figure 4. Comparison of phenotypic residual variance changes for different scenarios of missing observations after 35 d for residual feed intake. Full = no missing data; 10% = 10% data missing; 20% = 20% data missing; 30% = 30% data missing.
|
|
Missing Observations
The effects of missing observations during the test were also examined by randomly deleting FI observations at 5, 10, 15, 20, 25, and 30% after 35 d on test. Although 6 different scenarios of datasets with missing observations were generated and analyzed in this study, only 3 scenarios (10, 20, and 30%) of missing observation results are presented through Figures 1 to 4. These results show that random missing FI data after 5 wk (35 d) neither affects the conclusion drawn from the full data set nor has a significant effect on the accuracy of measuring RFI (r > 0.99, P < 0.01) if a repeated measures analysis is used. However, missing observations affect the model fit statistics, and only the RFI results are presented in Table 4 as an example. As the missing observations increase from 10 to 30%, the value of model fit statistics (Akaike information criterion corrected for finite sample and Bayesian information criterion) becomes larger (Table 4), indicating a poorer fit of the model to the data. This trend is similar for all traits conducted.
|
DISCUSSION
|
|---|
The current study indicated that the test length for ADG, DMI, FCR, and RFI can be shortened to 63, 35, 42, and 63 d, respectively, under the GrowSafe System when BW is measured weekly. Recommended test durations for ADG in the literature were 112 d (Franklin et al., 1987; Kemp, 1990; Brown et al., 1991), 84 d (Swiger and Hazel, 1961; Lui and Makarechian, 1993a,b), and 70 d (Archer et al., 1997).
The present results also indicate that an ADG test with a shortened duration (to 63 d) showed very little decrease in measurement accuracy as the test length was increased over the test period. The reasons why a shortened test length for ADG was recognized may partially be due to more frequent BW (weekly) measurements taken in this study. In many of the studies where longer test durations were recommended, biweekly BW measurements were taken. With more frequent weighings, one can obtain more weight gain information than with less frequent weighings, thereby shortening the test period. This has been pointed out by Archer et al. (1999) and Graham et al. (1999).
The other reason may be that with the repeated measures analysis used in this study, one could account for the correlations and covariances between different measures on the same individual, and each measurement can then use all information of other measurements made on the same individual through correlations among them. Thus, with relatively few number of measurements, one could obtain the same amount of information as more measurements in a longer test period that do not use the repeated measures analysis. However, the appropriate test duration that we have identified in this study is longer than 42 to 56 d (Archer and Bergh, 2000) for small and large biological types, respectively, and 56 d (Kearney et al., 2004). This might be because our BW measurements were taken on a weekly basis, whereas Archer and Bergh (2000) and Kearney et al. (2004) took BW measurements on a daily basis.
The present results also indicate that the optimal length of test for ADG is the determinant of the optimal test duration for feed efficiency traits. This agrees with Archer et al. (1997) and Arthur et al. (2004) who have pointed out that any improvement in the measurement accuracy on ADG by reducing the test length will automatically reduce the test duration of efficiency traits such as RFI and FCR. Therefore, a regular and shorter weighing schedule is critical to obtain accurate BW measurements for reducing the test duration for ADG and related efficiency traits.
The result in this study suggests that the test length for DMI could be shortened to 35 d, which is consistent with the finding by Archer et al. (1997) and much shorter than the 56 to 70 d recommended by Archer and Bergh (2000). This result also supports the idea that the measurement of FI is not the determinant for the shortening of the test duration for feed efficiency traits as reported by Archer et al. (1997). The reasons for accurate measurement of DMI in a shorter time may mainly be that the FI was measured daily with the automatic feeding system and the repeated measures analysis used in this study. The FI for each day provided a more accurate measure of DMI for the average FI of the week than that of ADG, which was measured weekly.
Traits of RFI and FCR are indexes of 2 component (feed intake and growth rate) traits for measuring the potential production efficiency of animals. The results in this study indicate that a 42-d test for FCR and a 63-d test for RFI would be a sufficient test length under the GrowSafe System when BW is measured weekly and a repeated measures analysis is used to determine optimum test duration. The relative changes of phenotypic residual variance (Table 4) reduced less than 2 and 1% with each additional week of test for FCR and RFI after 42 and 63 d, respectively. This implies that additional costs of maintaining test steers beyond 42 and 63 d for FCR and RFI, respectively, will not significantly improve the accuracy of the measurements. Therefore, under the GrowSafe System, increasing the days on test for the 2 traits beyond 42 and 63 d is not justifiable. The results of Pearson and Spearman correlation (Table 4 and Figure 5) are also supportive of this recommendation because both of these correlations have reached 0.90 (P < 0.01) for the recommended test duration for both traits. These results are comparable to the findings by Archer et al. (1997) but shorter than what they recommended, which was a 70-d test duration for both of these traits. The reasons for shorter test durations for FCR and RFI are similar to the reasons discussed for ADG and DMI.
View larger version (32K):
[in this window]
[in a new window]
|
Figure 5. Correlations between shortened tests and full test length for ADG, DMI, feed conversion ratio (FCR), and residual feed intake (RFI). Pearson = Pearson correlation; Spearman = Spearman correlation.
|
|
Random missing FI observations after 35 d in the test did not affect the conclusion drawn from the full data set (and these have been demonstrated through Figures 1 to 4) or have significant effects on accuracy of measuring RFI (r > 0.99, P < 0.01) when repeated measures analysis was used. The high correlation between missing observation data sets and the full data set (91 d) might be due to the repeated measures analysis where the analysis is able to capitalize on the correlations between different measurements on the same steer in their earlier records. The first 35-d measurements on an individual animal provided very reliable information for predicting the animal’s missing measurements later in the test (Craig and Schinckel, 2001; Wang and Zuidhof, 2004) because the analysis is able to make full use of the animals’ earlier information of measurements. However, Hebart et al. (2004) reached a similar conclusion without using repeated measures analysis. This study addressed the effect of missing data on overall accuracy using repeated measures methods, whereas Hebart et al. (2004) approached it using other Pearson correlation and t-test methods, and both studies have arrived at similar conclusions.
Reducing the duration of test has advantages in that one could increase the number of animals tested and reduce costs associated with the test. Usually, on a 80% grain 20% silage diet, the cost of 1 kg of gain was estimated at Can$1.22 (L. A. Goonewardene, Alberta Agriculture, Food and Rural Development, 2005, unpublished data). Although our study used steers, there is no reason to believe that the methodology and findings are not applicable to bulls, although bulls are known to grow faster than steers (Berg and Butterfield, 1976).
Four criteria were used to determine the optimal test duration in this study. These were variance reduction, relative changes of variances, and Pearson and Spearman rank correlations over time; all 4 approaches were complementary to one another. However, when correlations are calculated in overlapping periods, as is the case in this study, there is a tendency for the correlations to be greater toward the end of the test because of auto correlation (Wang et al., 2005). However, the decisions based on the reduction in variance and relative changes of variance over time are not affected by auto-correlated data. In addition, the rank correlation also provides a measure for a reliable decision for the test durations because it measures correspondence between ranks (Steel et al., 1997).
|
IMPLICATIONS
|
|---|
Average daily gain, dry matter intake, feed conversion ratio, and residual feed intake test duration could be shortened to 63, 35, 42, and 63 days, respectively, without significantly reducing the accuracy of the test when weight was measured weekly. As much as 30% of random missing data after 35 days in the test does not affect the accuracy of measurements for the traits studied. A mixed model repeated measures analysis with an appropriate covariance structure should be a good choice for this type of data to account for the auto correlations inherent in the data. These results have valuable practical implications for performance and feed efficiency testing in beef cattle.
|
Footnotes
|
|---|
1 This work was supported through grants 2000AB364, #BCRC 2002L030R, #ASAR AARI 2002L030R, and ABP/ACC bovine genome seed fund awarded to S. S. Moore through the Canada/Alberta Beef Industry Development Fund and Bovine Genome Project, Beef Cattle Research Council, and Alberta Beef Producers/Alberta Cattle Commission. The authors are thankful to staff at University of Alberta Kinsella Research Station for their support. 
2 Corresponding author: zhiquan{at}ualberta.ca
Received for publication December 12, 2005.
Accepted for publication April 25, 2006.
|
LITERATURE CITED
|
|---|
Archer, J. A., P. F. Arthur, R. M. Herd, P. F. Parnell, and W. S. Pitchford. 1997. Optimum postweaning test for measurement of growth rate, feed intake and feed efficiency in British breed cattle. J. Anim. Sci. 75:2024–2032.[Abstract/Free Full Text]
Archer, J. A., and L. Bergh. 2000. Duration of performance tests for growth rate, feed intake and feed efficiency in four biological types of beef cattle. Livest. Prod. Sci. 65:47–55.[CrossRef]
Archer, J. A., E. C. Richardson, R. M. Herd, and P. F. Arthur. 1999. Potential for selection to improve efficiency of feed use in beef cattle: A review. Aust. J. Agric. Res. 50:147–161.[CrossRef]
Arthur, P. F., J. A. Archer, and R. M. Herd. 2004. Feed intake and efficiency in beef cattle: Overview of recent Australian research and challenges for future. Aust. J. Exp. Agric. 44:361–369.[CrossRef]
Arthur, P. F., G. Renand, and D. Krauss. 2001. Genetic and phenotypic relationships among different measures of growth and feed efficiency in young Charolais bulls. Livest. Prod. Sci. 68:131–139.[CrossRef]
Basarab, J. A., M. A. Price, J. L. Aalhus, E. K. Okine, W. M. Snelling, and K. L. Lyle. 2003. Residual feed intake and body composition in young growing cattle. Can. J. Anim. Sci. 83:189–204.
Berg, R. T., and R. M. Butterfield. 1976. New Concepts of Cattle Growth. Univ. Sydney Press, Sydney, Australia.
Berg, R. T., M. Makarechian, and P. F. Arthur. 1990. The University of Alberta beef breeding project after 30 years—A review. Univ. Alberta Annu. Feeders’ Day Rep. 69:65–69.
Brown, A. H., Jr., J. J. Chewning, Z. B. Johnson, W. C. Loe, and C. J. Brown. 1991. Effects of 84-, 112-, and 140-day postweaning feedlot performance tests for beef bulls. J. Anim. Sci. 69:451–461.[Abstract]
Canadian Council on Animal Care. 1993. Guide to the Care and Use of Experimental Animals. Vol. 1. E. D. Olfert, B. M. Cross, and A. A. McWilliams, ed. Can. Counc. Anim. Care, Ottawa, Ontario, Canada.
Craig, B. A., and A. P. Schinckel. 2001. Nonlinear mixed effects model for swine growth. Prof. Anim. Sci. 17:256–260.
Franklin, C. L., W. V. Thayne, W. R. Wagner, L. P. Stevens, and E. K. Inskeep. 1987. Factors affecting gain of beef bulls consigned to a central test station. Bulletin 693, Agricultural and Forestry Experiment Station. West Virginia Univ., Morgantown.
Goonewardene, L. A., Z. Wang, M. A. Price, R.-C. Yang, R. T. Berg, and M. Makarechian. 2003. The effect of udder type and calving assistance on weaning traits of beef and dairy x beef calves. Livest. Prod. Sci. 81:47–56.[CrossRef]
Graham, J. F., B. K. Knee, A. J. Clark, and G. A. Kearney. 1999. The potential to shorten the feeding period when measuring the net feed conversion efficiency of cattle using automated feeding and weighing system. Proc. Adv. Anim. Breed. Genet. 13:488–491.
Hebart, M. L., W. S. Pitchford, P. F. Arthur, J. A. Archer, R. M. Herd, and C. D. K. Bottema. 2004. Effect of missing data on the estimate of average daily feed intake in beef cattle. Aust. J. Exp. Agric. 44:415–421.[CrossRef]
Kearney, G. A., B. W. Knee, J. F. Graham, and S. A. Knott. 2004. The length of test required to measure liveweight change when test for feed efficiency in cattle. Aust. J. Exp. Agric. 44:411–414.[CrossRef]
Kemp, R. A. 1990. Relationships among test length and absolute and relative growth rate in central bull tests. J. Anim. Sci. 68:624–629.
Kennedy, B. W., J. H. J. van der Werf, and T. H. E. Meuwissen. 1993. Genetic and statistical properties of residual feed intake. J. Anim. Sci. 71:3239–3250.[Abstract]
Littell, R. C., R. C. Henry, and C. B. Ammerman. 1998. Statistical analysis of repeated measures data using SAS procedures. J. Anim. Sci. 76:1216–1231.[Abstract/Free Full Text]
Littell, R. C., J. Pendergast, and R. Natarajan. 2000. Modeling covariance structure in the analysis of repeated measures data. Stat. Med. 19:1793–1819.[CrossRef][Medline]
Lui, M. F., and M. Makarechian. 1993a. Factors influencing growth performance of beef bulls in test station. J. Anim. Sci. 71:1123–1127.[Abstract]
Lui, M. F., and M. Makarechian. 1993b. Optimum test period and associations between standard 140-day test period and shorter test periods for growth rate in station tested beef bulls. J. Anim. Breed. Genet. 110:312–317.
Nkrumah, J. D., J. A. Basarab, M. A. Price, E. K. Okine, A. Ammoura, S. Guercio, C. Hansen, C. Li, B. Benkel, B. Murdoch, and S. S. Moore. 2004. Different measures of energetic efficiency and their phenotypic relationships with growth, feed intake, and ultrasound and carcass merit in hybrid cattle. J. Anim. Sci. 82:2451–2459.[Abstract/Free Full Text]
Steel, R. G. D., J. H. Torrie, and D. A. Dickey. 1997. Principles and procedures of statistics: A biometrical approach. 3rd ed. McGraw-Hill Companies Inc., New York, NY.
Swiger, L. A., and L. N. Hazel. 1961. Optimum length of feeding period in selecting for gain of beef cattle. J. Anim. Sci. 20:189–194.[Abstract/Free Full Text]
Templeman, R. J., L. W. Douglass, and B. A. Craig. 2002. NCR-170 FASS Mixed Model Workshop 2002. ADSA-ASAS-CSAS National Meetings, Quebec City, QC, Canada, July 24–25, 2002. Fed. Anim. Sci. Soc., Savoy, IL.
Wang, Z., J. A. Basarab, L. A. Goonewardene, D. H. Crews Jr., P. Ramsey, K. L. Lyle, N. French, E. K. Okine, and S. Moore. 2005. Optimizing test duration for net feed efficiency in commercial bulls. J. Anim. Vet. Adv. 4:825–830.
Wang, Z., and L. A. Goonewardene. 2004. The use of mixed models in the analysis of animal experiments with repeated measures over time. Can. J. Anim. Sci. 84:1–11.
Wang, Z., and M. J. Zuidhof. 2004. Estimation of growth parameters using a nonlinear mixed Gompertz model. Poult. Sci. 83:847–852.[Abstract/Free Full Text]
Wolfinger, R. D. 1996. Heterogeneous variance covariance structures for repeated measures. J. Agric. Biol. Environ. Stat. 1:205–230.[CrossRef]
This article has been cited by other articles:
|
|
|
|
 
P. F. Arthur, I. M. Barchia, and L. R. Giles
Optimum duration of performance tests for evaluating growing pigs for growth and feed efficiency traits
J Anim Sci,
May 1, 2008;
86(5):
1096 - 1105.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
INTRODUCTION
