The Effect of Sampling Variables on Sample Accuracy and Distribution Characteristics

Abstract

Three studies were conducted using 50 to 53-day-old broilers to determine the ability of a sample to estimate certain parameters of the group from which it was taken. The effects of sampling, processing, reduction to parts, sample size, sex, and genetic cross on the distribution characteristics of the sample were also evaluated.The range method of sample selection produced samples that estimated the replicate means of two genetic crosses, AA/AA and P/AA, with equal accuracy. The random method of sample selection estimated the replicate means of the two crosses with a different (P<.05) degree of accuracy. The three sex group means, males, females and mixed sex, were estimated with equal precision. Range size increases from ± 2 and ± 4% to ± 6% resulted in an increase (P<.05) in the accuracy of the sample to predict the replicate means, but increases in sample size from 10 to 20, 30, or 40 had no effect on accuracy.Distribution characteristics were evaluated to determine their relationship to sample accuracy. Genetic cross differences did not affect the coefficients of kurtosis (COK) from samples chosen using the two methods of sample selection. Both the range and random selection method produced samples whose COK were affected (P<.05) by the sex of the sample. The range selection method produced samples that were more peaked (P<.05) when the live birds were processed to carcasses and parts. As sample size increased, the COK increased (P<.05); but as range size increased, the COK was not affected.The AA/AA samples chosen by the random method were more skewed (P<.05) than those chosen by the range method. This resulted in the decrease in accuracy found when the genetic crosses were compared. The females had larger (P<.05) coefficients of skewness (COS) than males, and in most cases, the samples that were skewed to the greatest degree were more peaked. Sample size and range size did not produce changes in COS values, but the COS values were larger (P<.05) when the birds were processed and reduced to parts.Coefficients of variability (COV) values showed that samples selected by the range method of sample selection were less variable than those chosen by the random method. Variability was found to be affected by both COK and COS differences, but in these studies, COS seemed to have the most affect on COV.