Latent class growth analysis for ordinal response data in the Distress Assessment and Response Tool: an evaluation of state-of-the-art implementations

06/07/2021 ∙ by Jianhui Gao, et al. ∙ 0

In clinical psychology, longitudinal studies are often conducted to understand developmental trends over a period of time. Latent class growth analysis is a standard and popular statistical approach to identify underlying subpopulations with heterogenous trends. Several implementations, such as LCGA (Mplus), proc traj (SAS) and package lcmm (R) are specially designed to perform latent growth analysis. Motivated by data collection of psychological instruments over time in the largest cancer centre in Canada, we compare these implementations using various simulated Edmonton Symptom Assessment System revised (ESAS-r) scores, an ordinal outcome from 0 to 10, in a large data-set consisting of more than 20,000 patients collected through the Distress Assessment and Response Tool (DART) from the Princess Margaret Cancer Center. We have found that both Mplus and lcmm lead to high correct classification rate, but Proc Traj tends to over estimate the number of classes and fails to converge. While Mplus is computationally more affordable than lcmm, it has a limit of maximum 10 levels for ordinal data. We therefore suggest first analyzing data on the original scale using lcmm. If computational time becomes an issue, then one can group the scores into categories and implement them in Mplus.

READ FULL TEXT
POST COMMENT

Comments

There are no comments yet.

Authors

page 8

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.