The egg data from six studies published earlier were re-examined to consider the merits of prediction equations derived from one data set in estimating parameters from another. Of particular interest was an equation suggested by Hunt (1963), which predicted shell percentage on the basis of the inverse of specific gravity. All data used here were obtained from eggs of White Leghorn hens. Five of the studies were conducted at the Animal Research Centre; the sixth involved eggs obtained from a commercial egg grading station. Data sets ranged in size from 228 to 5,736 observations. In each case, a regression equation was derived from the entire data set; in several cases, equations were also determined for subsets.Both the slopes of the regression line and the correlations were found to vary considerably among data sets (slopes from -65 to -146 and correlations from -.53 to -.92). Much of the variation could be attributed to several factors, including date of sampling, duration of storage, integrity of the shell, age, and strain of bird. When these factors had been taken into account, reasonably consistent estimates were generally obtained. However, with one data set, it was impossible to obtain meaningful estimates because these factors had not been controlled. It is apparent from these results that prediction equations derived under a particular set of circumstances are not generally transferable to other data sets. (