Document Type



A three-parameter model was fitted to serial recumbent length data recorded at seven ages from one month to two years for 441 infants in the Fels Longitudinal Study who were born between 1931 and 1974. These infants included 80 parent-offspring pairs and 282 siblings. The aims of the study included the demonstration that a specific mathematical model could be fitted to these data and the analysis of the associations among the parameters of the model and between these parameters and adult status. The development of a suitable model was necessary to analyze fully the serial data recorded in several growth studies in the United States. The parameters of the model were the estimated recumbent length at 1 month (0A the intrinsic rate of growth in recumbent length (02) and the pattern of growth in recumbent length (03). A good fit was obtained for every individual with average residuals at each age and median mean square errors that were smaller than the measurement error. There were a few significant correlations between 0X or 02 with adult values for stature, weight, weight/stature2, head circumference or calf fat thickness. Even when significant, however, the parameters for growth during infancy explained only small proportions of the adult variance. The values for 03 were not correlated significantly with adult size for these variables. There were correlations of +0.5 to +0.7 between corresponding parameters for the same mathe­matical model fitted to recumbent length and to weight; the same mechanisms may control the rates of growth in these two measures of general body size. Comparisons were made between four groups of infants differing by year of birth. In boys, the mean 0! (estimated recumbent length at one month) was significantly smaller for the 1946 birth year group than for the 1958 birth year group. The mean 02 values were slightly, but significantly, larger for boys born about 1934 than for boys born about 1946 or 1958. These findings imply that there were some secular changes in estimated recumbent lengths at one month and in the intrinsic rates of growth in boys. Since the changes were small and inconsistent, and there were no corresponding changes in the girls, they probably reflected sampling bias. Parent­offspring correlations for the estimated parameters were non-significant with values from +0.06 to +0.2. Sib-sib correlations for 0j and 02 were significant (about +0.4), but those for 03 were not significant. The larger correlations for sib-sib pairs than for parent-offspring pairs probably reflect the common environment shared by the siblings when both the parents and their offspring were measured at the same ages during infancy. The model applied was considered appropriate because the fit was good; the model had only a few parameters that were almost independent of each other and each of these parameters could be interpreted biologically.