الفهرس 
Censored linear regression models for skewed and heavy tailed longitudinal data
AbstractOnly 14 pages are availabe for public view
 |  | from | 93 |  |  |
| | | from | 93 | | |
Abstract
Non-Normality models that depend on Skew-distributions instead of Normal ones are used to overcome the problems of modelling the heterogeneous data with the problems of asymmetry, multimodality and heavy tails.Arellano-Valle and Genton in (2010) introduced the Extended Multivariate Skew-t distribution that presents a very flexible class of distributions that differ in the parameterizations methodology yet all agreed on the same purpose of generating random data and deriving moments. Sophisticated likelihood structure and parameters estimation techniques in case of monotone and non-monotone missing data, are key issues while modelling longitudinal data.Diggle and Kenward in (1994) proposed a multivariate linear model for longitudinal data that is mixed with a logistic regression model for the DROP out.The SEM algorithm was proposed by Celuex and Diebolt in (1985), it is a stochastic version of the EM algorithm. In the SEM algorithm, the E-step is replaced by a stochastic version called the S-step, in which the missing data is generated using the conditional distribution of the missing data given the observed data. After that in the M-step, a maximization procedure is used to estimate the parameters of the likelihood of that complete data.The two steps will be repeated for sufficient number of iterations