The adaptive multi-domain cognitive training game is considered to be an effective approach for the cognitive rehabilitation of patients with cognitive decline. The efficacy of this training program is required to be examined, but the treatment effect for each individual usually differs substantively, which could be not easily interpreted with traditional methods. To identify distinct groups in trajectories of training outcomes, Latent class growth analysis (LCGA) and Growth Mixture Modelling (GMM) with the hlme function of the lcmm package were performed. Individuals with similar intercept and slope parameters are classified into the same subgroup.

1. Wrote R codes based on tidyverse package to conduct data wrangling of the MoCA scale and cognitive training score
2. Wrote R codes based on lcmm package to run LCGA, GMM-1 and GMM-2 models, visualized model output, and conducted model comparison via multiple parameters
3. Wrote R codes based on caret and nnet packages to train and test multinomial classification model