A comparative study on repeated measurements data in the presence of missing data


The occurrence of missing observations is nearly unavoidable in longitudi- nal studies where repeated measurements are taken over time on the same subject who may miss appointments or drop out during the study period. In this article, we use the Gaussian estimating objective function to esti- mate the regression and correlation parameters and handle missing data using multiple imputation. The estimation of these parameters is carried out simultaneously using the iterative Newton-Raphson algorithm and the expectation-maximization algorithm. These ideas are implemented using two real data sets and both algorithms showed comparable results with respect to the standard errors of the parameters of interest.

DOI Code: 10.1285/i20705948v16n2p410

Keywords: Correlation structure; Gaussian estimation; Imputation; Longitudinal data; Missing values

Full Text: pdf

Creative Commons License
This work is licensed under a Creative Commons Attribuzione - Non commerciale - Non opere derivate 3.0 Italia License.