ABSTRACT
Analysis of count event data such as mortality cases, were often modelled using Poisson regression model. Maximum likelihood procedures were used by using SAS software to estimate the model parameters of a Poisson regression model. However, the Negative Binomial distribution has been widely suggested as the alternative to the Poisson when there is proof of overdispersion phenomenon. We modelled the mortality cases as the dependent variable using Poisson and Negative Binomial regression and compare both of the models. The procedures were done in SAS by using the function PROC GENMOD. The results showed that the mortality data in Poisson regression exhibit large ratio values between deviance to degree of freedom which indicate model misspecification or overdispersion. This large ratio was found to be reduced in Negative Binomial regression. The Normal probability plot of Pearson residual confirmed that the Negative Binomial regression is a better model than Poisson regression in modelling the mortality data. The objective of this study is to compare the goodness of fit of Poisson regression model and Negative Binomial regression model in the application of air pollution epidemiologic time series study by using SAS software.
ABSTRACT
Many studies on adolescent smoking have been conducted in Malaysia, but very limited information is available on smoking amongst lower secondary school male students (Forms 1 and 2). We present data from a baseline study in Kota Tinggi District, Johor on the psychosocial factors, stages of smoking acquisition and susceptibility to smoking initiation and their relationship to adolescent smoking. The study is the first wave of a 3-year longitudinal study which was conducted from March 2007 to May 2009, aimed to describe the prevalence of smoking among students in the lower secondary classes. A three stage stratified sampling was performed to obtain a sample. The Bogus Pipeline Method was employed to confirm smoking status. Prevalence of smoking was 35.5%. Smoking prevalence among students of schools located in the Federal Land Development Authority (FELDA) settlement areas (42.9%) was two-fold higher than in the rural and town schools combined (20.29%). Using the Fagerstrom scale, 90% of current smokers had lower addiction to nicotine. Smoking was associated with peer smoking [OR, 4.19 (95% CI, 2.57-6.82)], having a brother smoking [2.17 (1.31-3.61)], parental smoking [1.73 (1.17-2.80)] and locality where respondents attend school [1.94(1.11-3.39)]. The study indicates that, the prevalence of smoking was high in all areas especially FELDA settlement areas. Measures such as teaching of skills to resist social pressure to smoke, establishment of peer support groups and involvement of parents in anti-smoking programs are recommended to curb the high prevalence of smoking among lower secondary school students in Kota Tinggi.