Growth Mixture Modeling for Sequential Growth Processes

Shige Song, University of California, Los Angeles
Bengt Muthen, University of California, Los Angeles

We re-visit the issue of “compensatory growth hypothesis”. The main idea is that some infants who experience substandard growth due to nutrition deficiency may grow faster than other children at a later age, and eventually catch up. We test this hypothesis using the data from the Cebu longitudinal study that records two developmental phases for each individual: the first developmental phase includes 12 bi-monthly records from age zero to age two, and the second developmental phase includes three records measured around age 7, 10 and 14. To test the compensatory growth hypothesis is to identify a subgroup in the sample that grows slower than others from age 0-2, but catches up from age 7-14. Being able to simultaneously identify the optimal number of subgroups that has qualitatively different growth trajectories and the growth trajectory within each subgroup based on empirical data is the major strength of growth mixture modeling technique.