Applied Multiple Poisson Regression for Dental Epidemiology Research

Abstract

In considering data as count data analysis especially for rare event and zero outcome, if the re-searcher analyzed data with inappropriate statistics, it will be very likely to get wrong evidence and errors leading to wrong conclusions. This study is a part of the analytical cross sectional study on “the association between daily pocket money and decayed, missing and filled teeth among children aged 12 Years in primary school at Khumuang District, Buriram Province”. The purpose of this study was to explore the application of multiple Poisson regression for dental epidemiology research. Samples were 299 students recruited by a systematic sampling method. An oral health survey form, and daily pocket money compila-tion forms and questionnaires for face-to-face interview were employed for data collection. It was found that there was 11.37% of over-dispersion problem and zero outcomes. Researcher selected the best statistics for analyzed the data by comparable mixed two processes between zero inflated negative binomial regression and zero inflated Poisson regression, and comparable mixed distributions between zero inflated negative binomial regression and negative binomial regression the results were found zero inflated negative binomial regression was appropriated more than zero inflated Poisson regression and better than negative binomial regression. There for advance statistics should be applied, for example correlation under Gener-alized Estimating Equation (GEE), Poisson data with nested data structures under generalized linear mixed model (GLMM), generalized additive mixed models (GAMM), multilevel with multiple Poisson regres-sion to developed quality of dental epidemiology and health science research.