Some considerations in the analysis of rates of change in longitudinal studies.

This paper discusses and compares several estimators of mean rate of change in unbalanced longitudinal data based on a model with randomly distributed regression coefficients across individuals. The estimators are unweighted and weighted means of these coefficients. The paper also evaluates commonly used variance estimates corresponding to the estimators. Results show that in situations of very slight imbalance, the choice of method is not critical. When imbalance is substantial, however, one should weight the regression coefficients by their estimated precision. An example using data from a nutritional study on premature neonates illustrates some issues encountered in the analysis of longitudinal clinical data sets.