Do Sophisticated Imputation Methods Improve Brain Age Prediction from Behavioral Data?

Learning objectives:

• Determine whether sophisticated imputation methods provide a predictive advantage over simpler methods for brain age regression modeling using sensitive linear models.
• Assess the relationship between imputation strategies and model performance as a function of missing data rates.
• Compare and analyze the distribution of the most influential features across imputed datasets, using the unmasked dataset as the reference baseline.

Introduction/Problem description:  Missing data is a challenge in clinical research, potentially reducing statistical influence. More sophisticated imputation methods are often employed over simpler methods. This study aims to empirically compare the impact of imputation strategies across clinically relevant missingness rates on the performance of a brain age predictor trained on clinical behavioral data. 

Full abstract available in the conference proceedings, available online and as a downloadable PDF.  See link below.