Linear mixed effects model for longitudinal data with non - random dropout /
Noha Ahmed Mohamed Youssef
Linear mixed effects model for longitudinal data with non - random dropout / النماذج الخطية ذات المؤثرات المختلطة للبيانات الطولية فى وجود قيم مفقودة نهائيا بصورة غير عشوائية Noha Ahmed Mohamed Youssef ; supervised Maged Osman , Thanaa Esmail - Cairo : Noha Ahmed Mohamed Youssef , 2004 - 89L ; 30cm
Thesis (M.Sc.) - Cairo University - Faculty Of Economics and Political Science - Department Of Statistics
Longitudinal studies represent one of the principal research strategies employed in medical and social researchThey are the most appropriate studies for studying individual change through timeNon - random dropout is a common phenomenon associated with this type of dataLinear mixed effects model has been used for fitting longitudinal data in the presence of non - random dropoutIt offers a powerful tool for analyzing longitudinal dataIt gives us information about within individual variation and between individuals variationStandard methods of maximum likelihood estimation are intractable in the current setting , especially when missing data mechanism is taken into consideration , which means incorporating a model for the dropout mechanismThis incorporation increases the number of parameters needs to be estimatedThe selection model proposed by Diggle and Kenward (1994) has been followed in this study to model the dropout processIn this thesis , the stochastic EM algorithm has been applied for the first time to obtain the maximum likelihood estimates of the linear mixed effects model parameters in the case of non - random dropout besides the maximum likelihood estimates of the parameters that control the dropout processSince the result of this algorithm is a Markov chain , three methods of monitoring convergence have been used to assess convergence , which are Gelman and Rubin's method (1982) , Yu and Mykland's method (1998) and Brooks' method (1998) The bootstrap method has been used to compute the standard errors for the parameters estimatesAll these methods are applied to two real data sets
Linear mixed effects model Longitudinal Data Missing data ; Non - random dropout The Expectation Maximization algorithm
Linear mixed effects model for longitudinal data with non - random dropout / النماذج الخطية ذات المؤثرات المختلطة للبيانات الطولية فى وجود قيم مفقودة نهائيا بصورة غير عشوائية Noha Ahmed Mohamed Youssef ; supervised Maged Osman , Thanaa Esmail - Cairo : Noha Ahmed Mohamed Youssef , 2004 - 89L ; 30cm
Thesis (M.Sc.) - Cairo University - Faculty Of Economics and Political Science - Department Of Statistics
Longitudinal studies represent one of the principal research strategies employed in medical and social researchThey are the most appropriate studies for studying individual change through timeNon - random dropout is a common phenomenon associated with this type of dataLinear mixed effects model has been used for fitting longitudinal data in the presence of non - random dropoutIt offers a powerful tool for analyzing longitudinal dataIt gives us information about within individual variation and between individuals variationStandard methods of maximum likelihood estimation are intractable in the current setting , especially when missing data mechanism is taken into consideration , which means incorporating a model for the dropout mechanismThis incorporation increases the number of parameters needs to be estimatedThe selection model proposed by Diggle and Kenward (1994) has been followed in this study to model the dropout processIn this thesis , the stochastic EM algorithm has been applied for the first time to obtain the maximum likelihood estimates of the linear mixed effects model parameters in the case of non - random dropout besides the maximum likelihood estimates of the parameters that control the dropout processSince the result of this algorithm is a Markov chain , three methods of monitoring convergence have been used to assess convergence , which are Gelman and Rubin's method (1982) , Yu and Mykland's method (1998) and Brooks' method (1998) The bootstrap method has been used to compute the standard errors for the parameters estimatesAll these methods are applied to two real data sets
Linear mixed effects model Longitudinal Data Missing data ; Non - random dropout The Expectation Maximization algorithm