Department of Agricultural Science, the State University of Western Parana, Marechal Cândido Rondon PR, (45) 3284-7912, Brazil.
Department of Poultry Science, The University of Georgia, Athens, GA 30602-2772, USA.
Poult Sci. 2018 Oct 1;97(10):3746-3754. doi: 10.3382/ps/pey217.
Experimental power is a measure of the ability of an experiment to detect differences between treatment means. Researchers design experiments and then calculate the probability that differences are simply due to chance, the null hypothesis. The objective of the analyses reported here was to determine the appropriate number of samples to demonstrate significant differences of various magnitudes from broiler chicken blood constituents. Over 800 samples were taken for a study of the effects of sample storage time, serum vs. plasma, light intensity, and fed vs. fasted birds on blood cholesterol, triglycerides, uric acid, glucose, total protein (TP), albumin, globulin, alanine aminotransferase (ALT), aspartate aminotransferase, gammaGT, creatinine, alkaline phosphatase, Ca and P. Various transformations increased the QQ plot R2 values from 0.000 to 0.149 or 0.00 to 17.62%. Most of the QQ plot R2 values were at or above 0.90. The 1/x2 transformation of blood P data showed the biggest increase in QQ plot R2 (0.846 to 0.995). The different standard deviations and coefficients of variation (CVs) found for each variable resulted in widely different numbers of replicates needed to detect differences in 2 treatment means. The extremes were glucose with a CV of 6.9% and ALT with a CV of 39.7%. For glucose, 15 replicates are needed to find a 10% difference in 97% of experiments; for ALT, 15 replicates would detect a 50% difference 91% of the time. The use of parameters such as cholesterol, glucose, TP, albumin, and globulin showed low CVs, indicating they may be considered as stable parameters. The lower CVs make it possible to find differences with a smaller number of replicates used in studies. As reported, the phosphorus values did not have a normal distribution of the data, so a transformation of these data could be an alternative to better discuss the results found.
实验能力是衡量实验检测处理均值差异的能力的指标。研究人员设计实验，然后计算差异仅由于偶然因素（零假设）的概率。这里报告的分析目的是确定适当的样本数量，以证明肉鸡血液成分的各种大小的显著差异。为了研究样本储存时间、血清与血浆、光照强度、空腹与喂食鸡对血液胆固醇、甘油三酯、尿酸、葡萄糖、总蛋白（TP）、白蛋白、球蛋白、丙氨酸氨基转移酶（ALT）、天门冬氨酸氨基转移酶、γGT、肌酐、碱性磷酸酶、Ca 和 P 的影响，采集了 800 多个样本。各种转换将 QQ 图 R2 值从 0.000 增加到 0.149 或 0.00 增加到 17.62%。大多数 QQ 图 R2 值在 0.90 或更高。血液 P 数据的 1/x2 转换显示了 QQ 图 R2 值的最大增加（从 0.846 增加到 0.995）。由于每个变量的标准偏差和变异系数（CV）不同，因此需要检测两种处理方法之间的差异的重复数量也大不相同。极端情况是葡萄糖的 CV 为 6.9%，ALT 的 CV 为 39.7%。对于葡萄糖，在 97%的实验中，需要 15 个重复才能发现 10%的差异；对于 ALT，在 91%的时间内，15 个重复可以检测到 50%的差异。胆固醇、葡萄糖、TP、白蛋白和球蛋白等参数的使用显示出较低的 CV，表明它们可能被视为稳定参数。较低的 CV 使得在研究中使用较少的重复来发现差异成为可能。如前所述，磷值的数据没有正态分布，因此对这些数据进行转换可能是更好地讨论所发现结果的一种替代方法。
