|
Abstract:
|
Margin of Error (MOE), or measure of the variability of the estimate due to sampling error exists in survey data. MOE is often complete ignored or not treated properly due to the fact the standard statistical textbooks on regression often pay very little attention to this aspect. Parameter estimates and confidence interval may suffer from serious biases if MOE is ignored. Bayesian approaches provide a flexible framework, as knowledge about the uncertainty of the covariates can be incorporated into the prior distribution. In this study, we formulated a Bayesian generalized linear mixed model to accommodate the uncertainty of the covariates. We presented the results in the context of methicillin-resistant Staphylococcus aureus (MRSA) infections, where rates among black are higher than among white persons. Since some factors related to socioeconomic status have been described to be associated with MRSA infection and colonization, we use data from American Community Survey (ACS) to explore to which extent the socioeconomic factors might explain racial variations in community-associated MRSA rates. We show how parameter estimates are obtained when accounting for MOE of ACS data.
|
