In this paper an algorithm called SEM, which is a stochastic version of the EM algorithm, is used to analyze multivariate skew-normal data with intermittent missing values. Also, a multivariate selection model framework for modeling of both missing and response mechanisms is formulated. By the SEM algorithm missing values of responses are inputed by the conditional distribution of missing values given observed data and then the log-likelihood of the pseudocomplete data is maximized. The algorithm is iterated until convergence of parameter estimates. Results of an application are also reported where a Bootstrap approach is used to compute the standard error of the parameter estimates.