The purpose of this study was to examine the potential impact of selected methodological factors on the validity of conclusions from reliability generalization (RG) studies. The study focused on four factors; (1) missing data in the primary studies; (2) transformation of sample reliability estimates; (3) use of sample weights for estimating mean score reliability and building confidence bands; and (4) differences between analyses of score reliability estimates and estimates of standard error of measurement. The research was a Monte Carlo study in which random samples were simulated under known and controlled population conditions. In the Monte Carlo study, RG studies were simulated by generating samples in primary studies, estimating the reliability of scores in these samples, and then aggregating the sample reliability estimates in the RG studies. In general, the results suggest that the use of Fishers z transformation of the reliability estimates provided a modest increase in the accuracy of the estimation of the population mean reliability. Although the statistical bias in point estimates of the mean reliability were very small across most conditions, the confidence bands obtained using the Fisher transformation were more accurate than the confidence bands obtained using the untransformed r. To refer to a metaphor developed by B. Thompson and Y. Vacha-Haase (2000), the RG chef needs to make sure that the ingredients are measured and prepared correctly to ensure that the RG sausage does not leave the reader with an unpleasant aftertaste. (Contains 6 figures, 23 tables, and 27 references.) (SLD)