robust longitudinal data analysis using a novel multivariate huberdistribution

This paper proposes a robust marginal linear model using a constructivemultivariate huber distribution for the longitudinal data analysis. it uses anunconstrained parameterization through the modified cholesky decomposition, andpresents a convenient technique to estimate the underlying covariance matrix, andallows subject-varying the tuning parameter. we present the estimation process byutilizing the exact likelihood function via the hamiltonian monte carlo algorithm. tohighlight the advantage of our model, we conducted a simulation study and reanalyzedan empirical application on economic studies.