000			01916 a2200253 4500
999			_c550 _d550
003			OSt
005			20220906184545.0
008			190213b ||||| |||| 00| 0 eng d
060			_aTHS-00083
100			_aDias, Jose Goncalves. _91715
245			_aFinite mixture models with applications to demography.
260			_cc2001.
300			_a61p.
500			_aThesis Report.
520			_aABSTRACT: This thesis intends to be methodological. We propose a general framework to incorporate unobserved heterogeneity in the statistical analysis that can be easily applied in demographic research. Before introducing the finite mixture framework, we revise the maximum likelihood estimation in the homogeneous case. Then, we extend the heterogeneous case, assuming and unknown number of components in the finite mixture for independent observations. Furthermore, we extend this framework to accommodate dependent observations following a Markov chain (finite mixture of Markov chains). Finally, a model defined for static (cross-sectional) and dynamic (longitudinal) variables is proposed. These models were implemented in MATALAB5.3. The models defined are illustrated, using the data from Brazil Demographic and Health Survey (BDHS) 1996, focusing on women's attitudes towards family planning and contraceptive use dynamics. Individual respondents (women) were segmented based on different attitudes reported towards family planning and contraceptive use dynamics. Key words : Finite mixture models, Latent class analysis, Segmentation techniques, Markov models, Contraceptive use dynamics
546			_aEng.
650			_aFinite mixture models. _91787
650			_aLatent class analysis. _91788
650			_aSegmentation techniques. _91789
650			_aMarkov models. _91790
650			_aContraceptive use dynamics. _91791
856			_uhttp://nhrc.gov.np/contact/ _yVisit NHRC Library
942			_2NLM _cTR