Variation In ADG Research Articles

208 Do prediction equations with high R-squared in low sample size studies have high predictive ability?

Abstract The use of R-squared as a measurement of prediction in regression is a common practice in animal nutrition studies. Generally, studies use data used to estimate the prediction equation to calculate the R-squared of the equation, which does not represent prediction. The objective was to evaluate the predictive ability of prediction equations using independent datasets. Data were simulated in based on previously reported prediction equations for ADG in nursery pigs: 152 + 195(SID_Lysine%). A completely randomized design with SID_Lysine% levels of 1.2, 1.3, and 1.4% was used. Data were simulated for three CVs representing low, moderate, and high variation in ADG, for low number of replicates per treatment (RepTreat; 4 to 10, every 2). Data were simulated and analyses performed 300 times for each scenario. Prediction equations were estimated for each scenario when significant effect of SID_Lysine% was followed by a significant orthogonal contrast testing the linear effect of SID_Lysine% (P-value< 0.05), and their R-squared were calculated (R2_training). The prediction equation was used to test the fit of the other datasets, and their R-squared were calculated and averaged for each prediction equation (R2_prediction). The relationship between R2_training and R2_prediction was assessed through Spearmen’s correlation and regression of R2_prediction on R2_training, across scenarios. The R2_training and R2_prediction averages were similar with 94.5% and 93.9%, 66.0% and 63.5%, and 46.1% and 42.7%, for low, moderate, and high CVs, respectively. The rank-correlation between R2_training and R2_prediction increased with CV: 0.23, 0.67, and 0.77, for low, moderate, and high CV, respectively. From regression analysis of R2_prediction on R2_training, as CV increased bias increased, with prediction equations with greater R2_training overestimating R2_prediction, and prediction equations with lower R2_training underestimating R2_prediction. These results indicate that in small sample sizes (≤10 per treatment), as CV increases, models with high R-squared are overestimated, and are not as predictable.

Association of preweaning and weaning serum cortisol and metabolites with ADG and incidence of respiratory disease in beef cattle.

The objectives of this experiment were to determine the association of circulating cortisol, lactate, and glucose at, and prior to, weaning with ADG and incidences of bovine respiratory disease (BRD) in beef cattle. A blood sample was collected approximately 3 wk prior to weaning and at weaning from genetically diverse steers and heifers ( = 451). Cattle were weighed periodically throughout the study and ADG was calculated for the preweaning period (152 ± 15 d), the receiving period (45 d postweaning), the finishing period (200 d), and total postweaning ADG. Incidences of BRD were recorded and analyzed as a binary trait. Lung lesions were recorded at slaughter. Preweaning serum cortisol concentrations were positively associated ( = 0.040) with receiving ADG and explained 0.74% of the variance of receiving ADG. Preweaning glucose concentrations were positively associated ( < 0.001) with preweaning ADG and negatively associated with receiving ( = 0.003), finishing ( = 0.008), and total postweaning ADG ( = 0.002) and explained 2.0% of the variance in total postweaning ADG. Variation in preweaning serum glucose concentrations could be indicative of variation in milk consumption, and therefore indicate calves receiving less milk grow slower prior to weaning, but experience compensatory gain postweaning. Cattle that were diagnosed with BRD ( = 130) grew slower during the receiving phase ( = 0.004), but total postweaning ADG was not different from cattle not diagnosed with BRD ( = 0.683). Additionally, cattle that were diagnosed with BRD in the feedlot tended ( = 0.062) to have slightly lower preweaning serum glucose concentrations. Using a logistic regression analysis, none of the serum variables measured at or before weaning were predictive of developing BRD ( > 0.180). Weaning serum glucose concentrations tended to be predictive of the presence of lung lesions at weaning ( = 0.060). These data indicate that glucose measured early in life is associated with growth rate, and could indicate that carbohydrate metabolism could contribute to variation in ADG.

Non-genetic sources of variation and temporal variability in growth and feed efficiency traits among phylogenetically distinct clusters of indigenous chicken in Kenya.

This study aims to investigate the influence of non-genetic factors on feed efficiency in indigenous chicken. Residual feed intake (RFI), residual gain (RG) and residual intake and gain (RIG) were used as measures of feed efficiency. Feed intake and body weight data were collected on 107 experimental birds on a daily and weekly basis, respectively, from ages 11 to 20weeks. A general linear model was fitted to determine the effect of sex, cluster, genotype and hatch group on mean performance and to assess temporal variation across clusters. The overall mean performance for daily gain (ADG), daily feed intake (ADFI), weekly metabolic body weight (MBW), RFI, RG and RIG was 10.38g/day, 133.01g/day, 164.12g/day, 0.00 (±14.23), 0.00 (±1.83) and 0.00 (±14.64), respectively. Sex significantly influenced variation in ADG and RG while hatch group influenced all traits except ADFI. Cluster and genotype had no effect on the traits. Interaction between sex and cluster significantly influenced ADFI, RFI and RIG. There was a significant temporal variation within and among clusters resulting in re-ranking of the phylogenetic groups in efficiency across the test period. Results indicate that growth and feed efficiency traits are influenced by non-genetic factors which should be accounted for, to reduce bias and improve accuracy of performance evaluations in the indigenous chicken.

Evaluating critical factors to the economic feasibility of semi-intensive pig rearing in western Kenya

The purpose of this research is to assess how season, ADG, opportunity costs of farm-grown feeds, pig weight, and butcher price variation impact the economic potential of semi-intensive pig rearing. We developed a unique algorithm that emulates least-cost pig feeding and used it to assess the impact of the aforementioned factors on farmers' maximum revenue and profit potential when pigs are sold to local butchers in western Kenya. When considered as independent factors influencing feed costs to grow a pig to a market weight of 30 kg, variation in ADG, opportunity cost of feed, and weaning season resulted in feed cost differences of up to 982, 947, and 379 Kenyan shillings (KES), respectively. The variation in revenues attributable to butcher or butcher negotiation and seasonal variance of butcher prices for a 30 kg pig was 744 and 225 KES, respectively. Feed items most commonly chosen for least-cost feed rations were small dried fish, cooked ground maize, whole maize, millet, cassava foliage, sweet potato vines, bone meal, avocado, and mango. Smallholder farmers who can feed pigs to reach higher ADG, have lower opportunity costs of feeds and/or who effectively bargain with butchers can benefit from semi-intensive pig rearing. Farmers without access to at least some zero-cost feeds and farmers with opportunity costs of feeds exceeding 50 % of the market price will not earn positive returns from semi-intensive pig rearing.

Performance of stocker cattle grazing 2 sorghum-sudangrass hybrids under various stocking rates

<h2>ABSTRACT</h2> Summer annual forages are a practical roughage source for stocker cattle operations because of their high yield potential. A 2-yr study was conducted to compare the performance of steers grazing 2 different varieties of sorghum-sudangrass hybrids at different stocking rates. Brown midrib (BMR) and photoperiod-sensitive (PS) sorghum-sudangrass hybrids were selected based on differences in nutritional value and yield. Twelve experimental pastures (6 per variety) were established and grazed by steers (231 ± 4 kg) for 84 d each year at stocking rates ranging from 2.33 to 4.10 animal units per hectare per 84 d (3.6 to 6.4 steers/ha). Initial (0 d), intermediate (28 and 56 d), and final (84 d) cattle BW was used to calculate ADG and gain per hectare. Forage samples were also collected at 28-d intervals to assess quality and quantity of standing forage. Regression analysis was used to compare cattle production on BMR and PS forage varieties. A curvilinear model with stocking rate and forage variety best fit the data, accounting for 73 and 72% of the variation in ADG and gain per hectare, respectively. At light to moderate stocking rates, ADG and gain per hectare were greater for the BMR; whereas the PS maintained BW gain at greater stocking rates. Our results indicate that cattle grazing these sorghum-sudangrass forages were capable of gaining over 1 kg/steer per day and up to 437 kg/ha over an 84-d grazing season, but response was dependent on forage variety and stocking rate of experimental pastures.

ALPHARMA BEEF CATTLE NUTRITION SYMPOSIUM: Predictability of feedlot cattle growth performance1

Predicting performance is vital to management and marketing decisions in commercial feedlots. Agreement between performance predicted from NE equations or empirical regression relationships and actual performance is generally very good, suggesting that factors affecting performance by finishing cattle are fairly well documented. The challenge for feedlot managers is to predict performance with limited information at the start of the feeding period. Data on sex and initial shrunk BW (ISBW) are typically available when cattle start on feed. Relationships between ISBW, sex, and performance were evaluated using 3,363 pen records collected over 4 yr from 3 commercial feedlots in the Texas Panhandle. Mixed-model regression was used to account for random effects of feedlot × season × year and fixed effects of ISBW (range = 227 to 451 kg), sex (steer or heifer), and ISBW × sex (P < 0.10 for all variables evaluated). Previously developed equations indicated that with intercept and slope adjustments for sex, ISBW accounted for 76 and 84% of the variation in DMI and final shrunk BW (FSBW), respectively. Similarly, newly developed regression equations that included ISBW, sex, and ISBW × sex accounted for 46 and 81% of the variation in ADG and HCW, respectively. Initial BW was negatively related to G:F (R(2) = 0.22). Including early DMI data (DMI from d 8 to 28) increased R(2) and decreased prediction error for DMI, indicating that updating predictions with interim intake data might prove beneficial. An independent data set (781 lots of steers and heifers) collected during 1 yr from 2 Texas Panhandle feedlots was used to validate equations developed with the larger database. Dry matter intake predicted from ISBW and sex accounted for 69% of the variation in observed DMI (SE of prediction = 0.47; mean bias = 0.42 kg). Predicting DMI with ISBW, sex, and DMI from d 8 to 28 of the feeding period increased r(2) to 0.76 and slightly decreased the SE of prediction (0.42 kg), but the equation had a strong linear bias (-0.174; P < 0.001). The r(2) values for regression of observed on predicted ADG, G:F, FSBW, and HCW were 0.37, 0.08, 0.74, and 0.73, respectively, with positive mean bias (underprediction for all equations). Average daily gain calculated with NE equations from predicted DMI (ISBW and sex equation) and predicted FSBW had a similar r(2) (0.38) but less mean bias (-0.08 kg) than ADG predicted directly from ISBW and sex. Adjustments to equations for animal type, health, and management effects would likely improve predictions. Nonetheless, results suggest that predicting performance from initial BW with adjustments for steers vs. heifers should have considerable utility in practical settings.

Maternal and social genetic effects on average daily gain of piglets from birth until weaning1

The aim of this study was to investigate whether there is heritable social variation in ADG from birth until weaning in piglets. Nursing and the establishment of teat order are sources of social interaction among suckling piglets nursed by the same sow. If a heritable social effect is present, but ignored, the selected animals might be the most competitive ones with negative effects on growth of their group mates, resulting in less response to selection than expected. The social interaction model was extended with a maternal component to estimate genetic maternal and social effects. Four different animal models were compared: a basic model with a direct heritable effect only; a social model accounting for direct and social heritable effects; a maternal model with a heritable maternal effect in addition to the basic model; and a social-maternal model accounting for direct, social, and maternal heritable effects. Estimates of direct, maternal, and social heritability were 0.07, 0.06, and around 0.0007 (not significantly different from zero, SE = 0.0005), respectively. Total heritable variance, including direct, social, and maternal heritable variance and their covariances ranged from 0.07 to 0.15 of the phenotypic variation. Both maternal models were significantly better than equivalent nonmaternal models (P <or= 0.005). The social model was not significantly better than the basic model (P = 0.102), and the social-maternal model was also not significantly better than the maternal model (P = 0.486). There was no evidence for heritable social effects among piglets in a group. The generally used maternal model fit the data as well as the social-maternal model. Sufficient cross-fostering is needed to partition social and maternal variation.

Application of broken-line analysis to assess floor space requirements of nursery and grower-finisher pigs expressed on an allometric basis1

Few issues in swine production are as complex as floor space allowances. One method for pork producers to calculate floor space allowance (A) is to convert BW into a 2-dimensional concept yielding an expression of A = k * BW(0.667). Data on ADG, ADFI, and G:F were obtained from published peer-reviewed studies. Five data sets were created: A = grower-finisher pigs, fully slatted floors, and consistent group size; B = grower-finisher pigs and fully slatted floors (group size did not need to be consistent); C = grower-finisher pigs, partially slatted floors, and consistent group size; D = grower-finisher pigs, partially slatted floors (group size did not need to be consistent); and E = nursery pigs, fully slatted or woven wire floors (group size did not need to be consistent). Each data set was analyzed using a broken-line analysis and a linear regression. For the broken-line analyses, the critical k value, below which a decrease in ADG occurred, varied from 0.0317 to 0.0348. In all cases the effect of space allowance on ADG was significant (P < 0.05). Using the linear analyses based on data with k values of < 0.030, the critical k values for the 4 grower-finisher data sets did not differ from those obtained using the broken-line analysis (0.0358 vs. 0.0336, respectively; P > 0.10); however, none of the linear regressions explained a significant proportion of the variation in ADG. The slopes for the nonplateau portion of the broken-line analyses based on percent values varied among data sets. For every 0.001 decrease in k (approximately 3% of the critical k value), ADG decreased by 0.56 to 1.41%, with an average value of 0.98% for the 5%-based analyses. The use of an allometric approach to express space allowance and broken-line analysis to establish space requirements seem to be useful tools for pig production. The critical k value at which crowding becomes detrimental to the growth of the pig is similar in full- and partial-slat systems and in nursery and grower-finisher stages. The critical point for crowding determined in these analyses approximated current recommendations to ensure the welfare of pigs.

The Cornell Net Carbohydrate and Protein System model for evaluating herd nutrition and nutrient excretion

The Cornell Net Carbohydrate and Protein System (CNCPS) is a mathematical model that estimates cattle requirements and nutrient supply based on animal, environmental, and feed compositional information in diverse production situations. Predicted animal requirements account for different physiological states (lactation, pregnancy, and growth), body reserves and environmental effects. The CNCPS uses feed carbohydrate and protein degradation and passage rates to predict extent of ruminal fermentation, microbial protein production, post-ruminal absorption, and total supply of metabolizable energy and protein to the animal. The CNCPS has been used successfully on beef and dairy cattle farms to evaluate and formulate rations. In an evaluation with individually fed growing cattle, the CNCPS accounted for 89% of the variation in ADG with a 7.4% underprediction bias. When the CNCPS was evaluated with data from individual dairy cows where the appropriate inputs were measured and changes in energy reserves were accounted for, the CNCPS accounted for 90% of the variation in actual milk production of individual cows with a 1.3% bias. The model accounted for 76% of the variation in individual cow milk production with an 8% underprediction bias when energy was first limiting in high producing cows, and accounted for 84% of the variation with a 1.1% overprediction bias when protein was first limiting.

Predicting growth in angus bulls: the use of GHRH challenge, insulin-like growth factor-I, and insulin-like growth factor binding proteins.

Plasma IGF-I, IGF binding protein-2 (IGFBP-2), and IGFBP-3 were quantified in growing Angus bulls (n = 56) to determine their relationship with postweaning growth and carcass ultrasound measurements. In addition, GH response to GHRH challenge (area-under-the-curve GH [AUC-GH) was determined for each bull as part of a previous study. Blood was collected by jugular venipuncture at the start of a 140-d postweaning growth performance test and at 28 d intervals for plasma IGF-I determination by RIA. Plasma IGFBP-2 and -3 content was measured at the start of the study, on d 70, and d 140 by Western ligand blotting. Individual weights and hip heights were measured every 28 d during the study and carcass longissimus muscle area, intramuscular fat percentage, and carcass backfat were estimated by ultrasound on d 140. Greater plasma IGF-I at the start of the performance test was associated with reduced postweaning ADG and increased longissimus area. Throughout the performance test period, the correlations between plasma IGF-I and hip height were consistently positive, ranging from 0.10 to 0.38, but the correlations between ADG and IGF-I varied from -0.32 to 0.31. Age-adjusted d-1 plasma IGFBP-2 was related to ADG during the performance test, explaining nearly 30% of the variation in ADG. A model combining weaning age, IGFBP-2, and AUC-GH showed a strong relationship with ADG (R2 = 0.40). Plasma IGFBP-2 and -3 were not related to carcass characteristics, and IGFBP-3 was not related to growth rates. This study provides additional evidence for the variable relationship between plasma IGF-I and growth rates in cattle. A significant positive relationship between plasma IGFBP-2, AUC-GH, and postweaning ADG warrants further investigation.

Application of the Cornell Net Carbohydrate and Protein model for cattle consuming forages.

Accurate prediction of forage biological values and performance with animals fed forages requires accurately accounting for factors that influence animal requirements and feedstuff utilization. The Cornell Net Carbohydrate and Protein System (CNCPS) is an application model that uses a combination of mechanistic and empirical approaches to account for the effects of variation in animal factors and feed carbohydrate and protein fractions on animal performance. Thus, accurate animal and environmental descriptions, DMI, feed carbohydrate, and protein fractions and their digestion rates are required inputs. In 25 growth periods with calves fed high-forage diets, the CNCPS accounted for 74, 81, and 83%, respectively, of the variation in ADG predicted to be supported by the ME, metabolizable protein, and essential amino acid intake, the first-limiting of all three accounting for 81% of the variation with a -1% bias. Thus, the CNCPS can be used to accurately describe forage quality and the effects of changes in forage composition on animal performance. The model was sensitive to variations in NDF, CP, protein solubility, NDF and starch digestion rates, feed and microbial amino acid composition, maintenance protein requirement, body protein amino acid content, and the coefficient of efficiency of use of absorbed protein. Analysis of several trials indicates an improved efficiency of ME use with improved amino acid balances. Uses of the CNCPS discussed include interpreting, planning and applying research, teaching, developing tables of requirements and biological values for feeds, complex nutritional accounting, and predicting performance and profits.

Animal model estimation of genetic parameters and response to selection for litter size and weight, growth, and backfat in closed seedstock populations of large white and Landrace swine2

Records from 2,495 litters and 14,605 Landrace and Large White pigs from two farms, but established from the same base population and run as replicated selection lines, were analyzed. Selection within herd was on estimated breeding values weighted by economic values. Animal models and REML procedures were used to estimate genetic, phenotypic, and environmental parameters for the number of pigs born alive (NBA), litter weight at 21 d (LW), average daily gain form approximately 30 to 104 kg (ADG), and backfat thickness adjusted to 104 kg (BF). Random animal genetic effects (o), permanent (NBA and LW) or litter (ADG and BF) environmental effects, maternal genetic effects (m), and the covariance between o and m were sequentially added to the model. Estimates of total heritability calculated from all data (ht2 = sigma o2 + 1/2 sigma m2 + 3/2 sigma om) ranged from .01 to .14 for NBA, from .18 to .22 for LW, from .23 to .34 for ADG, and from .40 to .50 for BF. Maternal genetic variance was from 2.4 to 3.8% of phenotypic variance in NBA, from 1.2 to 3.6% in LW, from .5 to 1.5% in ADG, and from 1.9 to 3.4% in BF. The correlation between o and m was -.07 for NBA, -.25 for LW, -.34 for ADG, and -.26 for BF. Permanent environmental effects explained from 16 to 17% of total phenotypic variation for NBA and from 1.6 to 5.3% for LW. Approximately 7% of the variation in ADG and 5% in BF was due to litter environmental effects.(ABSTRACT TRUNCATED AT 250 WORDS)