Decomposing design effects: Estimation of intra-class correlation in the selected Philippine household surveys

• Main Authors: Bondoc, Bernadette Marie M., Bondoc, Maria Urielle L.
• Format: text
• Language: English
• Published: Animo Repository 2009
• Subjects: Household surveys > Philippines; Sampling (Statistics); Estimation theory; Correlation (Statistics); Statistics and Probability
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/5221
• Institution: De La Salle University

This thesis presents a comprehensive decomposition of design effects into its components, particularly weighting and clustering from selected household survey in the Philippines, particularly, Family income and Expenditure Survey (FIES) for different years and Philippine Labor Force Survey (LFS). The expected design effect for each chosen variables, which is 3.05, is decomposed into its components specifically weighting and clustering. Design effects due to both components are then estimated. The clustering effect have shown an immense contribution on the overall design effect due to the high degree of clustering and the rate of the intra-class correlation employed in the sample design. The estimation of the intra-class correlation leads to the formulation of the implication involving the estimation of sample sizes as well as the optimal number of primary sampling units (PSU) and secondary sampling unit (SSU) under the consideration of a cost effective sample design. An increase in sample design would lead to an increase in the number of PSU and SSU. In this paper, the sample size employed in FIES and LFS is increased to 3,839 households to compensate for the design effect and intra-class correlation. This is converted to 19,195 individuals for the reason of the surveys utilized being multi-purpose in nature. In order to verify if the estimates have attained precision requirement, the margin of error with proportion as estimator and coefficient of variation with mean as estimator are determined. The following results would give a notion regarding the appropriate design in Philippine Household Surveys.