Accuracy of Growth Model Parameters: Effects of Frequency and Duration of Data Collection, and Missing Information

摘要

This study was done to compare the accuracy of prediction of growth parameters using the Gompertz model when (1) data was collected infrequently, (2) data collection was truncated, and (3) data was missing. Initial growth rate and rate of decay were reduced by half when the model was fitted to data collected biweekly compared to data collected weekly. This reduction led to an increase in age of maximum growth and subsequently over-predicted the asymptotic body weight. When only part of the growth duration was used for prediction, both the initial growth rate and rate of decay were reduced. The degree of data truncation also affected sexual dimorphism of the parameters estimated. Using pre-asymptotic data for growth parameter prediction does not allow the intrinsic efficiency of growth to be determined accurately. However, using growth data with body weights missing at different phases of the growth curve does not seem to significantly affect the predicted growth parameters. Speculative or diagnostic conclusions on intrinsic growth should be done with data collected at short intervals to avoid potential inaccuracies in the prediction of initial growth rate, exponential decay rate, age of maximum growth and asymptotic weight.