Detection and characterization of differences in plasma corticoid response to treatments

Ten experiments involving 3013 observations and 103 animals were studied to characterize variation associated with bovine plasma corticoid concentrations (ng/ml). Total corticoids, corticosterone, or cortisol were measured by competitve protein binding assay. Animals were noncystic or cystic, and measurements were made to establish basal concentrations or to estimate response to stimulation by adrenocorticotrophic hormone injection. Data were analyzed by least squares with the nested model: treatment regimen, animal in treatment, sample in animal in treatment, and sampling time to the highest order of regression (up to seventh) which was statistically significant (P less than .05). Estimates of error variances ranged from .31 for basal cortisol in cystic animals to 233.27 for total corticoids in noncystic, stimulated animals. Coefficients of variation among experiments ranged from 33 to 87% with mean 61%. In most analyses, estimates of experimental error were large indicating need for careful attention to experimental design to insure efficient least-cost experimentation. Use of the power function for nested analysis of variance to accomplish this is demonstrated.