ANDREI RAMANI, 100941024 (2011) PENGGUNAAN REGRESI LOGISTIK MULTILEVEL PADA DATA BERBASIS SURVEI (Studi kasus analisis kepesertaan KB pada pria dengan menggunakan data SDKI 2007). Thesis thesis, UNIVERSITAS AIRLANGGA.

Preview	Text (ABSTRAK) gdlhub-gdl-s2-2012-ramaniandr-22505-tkm171-k.pdf Download (308kB) | Preview
	Text (FULL TEXT) gdlhub-gdl-s2-2012-ramaniandr-18205-tkm1711.pdf Restricted to Registered users only Download (2MB) | Request a copy

Abstract

Multilevel analysis is a method of analysis used for the data which has hierarchial structure, with this analysis the researcher can see the effect on the individual level and macro level simultaneously (Duncan et al, 1998). The general objective of the study is to analyze the use of multilevel logistic regression on survey-based data about male family planning participation using IDHS 2007 data. Specific purposes are: 1) applying multilevel logistic regression analysis on data IDHS 2007 using 2-levels model, individuals and provincial 2) analyze the different variations on the individual level and provincial level, 3) analyze the value of the ICC, MOR, IOR, and R2MZ. This research was unobstrusive design using secondary IDHS 2007 data. 2-level multilevel logistic regression analysis was used in this research. Adaptive Gaussian Quadrature used in parameter estimation with the use of STATA software, supported with GLLAMM package (General Linear Latent and Mixed Models). Results showed the decision-making in male family planning participation predominantly influenced by variables at level-1 (individuals) which are knowledge and religion (Moslem - Non Moslem). Calculation of ICC, MOR, LR tests, and R2MZ revealed that the variation between individuals is greater than inter-provincial variation. Examination on multilevel logistic regression model through changes in condition number indicates that the fittest model in the analysis of male family planning participation is individual model (model 1). Procedures for fit model checking in multilevel logistic regression models using various indicators (ICC, MOR, R2MZ, and the likelihood ratio test) is a stage that can not be carried out step by step since the development of multilevel statistical software to this day has not been able to evaluate the multilevel regression models logistics as a whole. The use of these indicators can not be avoided because until now there is no standard agreement in fit model checking in multilevel logistic regression model.

Actions (login required)

View Item