Imputation Methods Approach to Clinical and Life Science Research Data Sets

Abstract

The missing values or items were aroused from various factors like lost to follow-up, critical illness, severe adverse drug reaction, due to death and some other accidental causes, and termination of clinical trial during the study period. In this intervention, the researcher is unable to draw the drug decision-making based on the negative predicted values. In medical research domain, the missing value imputation is a necessary footstep to extrapolate accurate scientific evidence (clinical data sets), knowledge of drug distribution, developing vaccine trails, automatic disease diagnosis, formulation of drug and induction of clinical trial, etc. Missing values may resort the final results, and the underlying data missing mechanism may deleteriously cause a biased statistical analysis and decrease the level of precision. Therefore, we need to be focused or formulate appropriate advanced statistical imputation models at larger extent. An advanced model will be necessary to predict the true predicted values at greater accuracy before testing the null hypothesis (H₀) at population or sample level. In this vein, the present chapter discusses various advanced imputation model for predicting missing values (least variance method, Gaussian mixture, KNN, and ARIMA models). An overall advanced imputed models were demonstrated by real data sets and diagnostically models were tested by suitable test statistics.