Repeatabilities of blood constituents were calculated for 104 Angus heifers on two separate experiments fed adequate, protein-deficient, energy-deficient, or both protein- and energy-deficient diets. Four statistical methodologies were compared including analysis of variance, principal component (structural) analysis based on the sample covariance and sample correlation matrix, and maximum likelihood. Of 12 blood constituents tested only seven were considered sufficiently important to be included in the analysis. These blood constituents included blood urea nitrogen (BUN), creatinine (Creat), alkaline phosphatase (Alk Phos), total protein (T Prot), total bilirubin (T Bil), cholesterol (Chol) and Iron (Fe). If the standard linear model assumptions were met for heifers on the adequate diet, the estimators appeared to be quite similar for both years except when the correlation coefficient was relatively small. If the assumption of homogeneity of the variance-covariance matrix (compound symmetry) was relaxed, the structural analysis method based on the sample correlation matrix appeared preferable. However, when combining all diets, the maximum likelihood methodology was preferred. Among the specific blood constituents, Alk Phos had the highest repeatability, not only for the heifers on the adequate diet, but also for heifers on other treatments in both years. Repeatabilities for T Prot appeared to be the most consistent over all rations in both years. Repeatability estimates for Fe were high and relative rankings were consistent for both years, while repeatabilities for the other variables were either low and(or) inconsistent.