Statistics

In this study an attempt has been made to describe the analysis of epidemiological study by Bayesian methods and apply this methodology to district level child mortality of Uttar Pradesh to assign rank to each district for rural and urban separately. The specific objectives of this study are to analysis of epidemiological study by use of fixed effect modeling and random effect modeling in Bayesian setup. To assign rank to each district by this suggest applying strategies to reduce child mortality in those district for those ranks are poor in Uttar Pradesh. The modified retrospective cohort study design used here. For fixed effect modeling beta-binomial modeling approach is used and for random effect modeling logit link function is used. The posterior estimates came in both cases under squared error loss function.

One of the main problems in sampling survey is the optimal allocation of resources. The solution of this problem is rather arbitrary due to the fact that no best allocation is defined. In this model, the allocation problem was to find the allocation of a sample to strata which minimizes cost of investigation. The idea of optimal allocation under a multivariate stratified sampling based on an alternative approach earlier worked on by Diaz- Garcia and Ranos– Quiroga was applied. The matrix of the variance-covariances of the vector of the stratified variables was obtained. An emperical data from a household survey conducted in Ogun state was used. The frame consisted 880,970 households in the twenty local government areas (LGA) of Ogun state. Each of the 20 LGAs that made up the state was seen as a stratified cluster. Post stratification in this study ensured that some variables that are suitable for stratification was achieved after selection of the sample. The four characteristics of interest were occupation, income, household size and educational level. The study transformed all variables in the stratified sampling plan by using dummy variables ranging from 1-3. The estimates used in the computation were calculated using statistical software Splus. The post stratification estimator ?st does not have the same variance as the stratified sample mean. Stratification produced a gain in precision in the estimation of characteristics of the household survey. The results of the estimates in the study revealed that proportionate allocation can lead to smaller sampling variances than for simple random samples of the same size.