Exploring missing data using adaptive LASSO regression imputation in relation to parkinson’s disease

Abstract

Parkinson’s disease (PD) belongs to a class of chronic disorders that has degenerative neurological symptoms. In the clinical trials, different results falling in the areas of binary, ordinal, and continuous are analyzed to detect manifestation of the symptoms of this disease. A global test statistic is used to comprehensively evaluate the impact of all sorts of results. However, this disease predominantly faces the challenge of missing data that arise in the clinical results for varied reasons such as dropout, death, etc., therefore, imputation of such missing data must be carried out before conducting an intent-to-treat analysis. In fact, accuracy in data pertaining to disease progression may not be possible through statistical analysis without application of an appropriate mechanism that effectively handles missing data. In the p[resent paper, an Adaptive LASSO Imputation method has been proposed with its foundational basis on item response theory so that multiple imputations can be performed while dealing with multiple sources of correlation. The Root Mean Square Error (RMSE) formula was applied to evaluate the precision of each imputation method. The obtained results prove the better performance analysis of the proposed technique over all the known different algorithms.