Comparison of time series regression models in environmental epidemiology for diseases modeling

Aim: In this study, 3 different outcome variables of Asthma, COPD, Pneumonia and Pte diseases were modeled by using GAM, DGDOM and GAMAR methods and the best model was investigated by comparing the performance results of the methods. Material and Methods: The data was obtained from 3 hospitals under the General Secretariat of Gaziantep province public hospitals retrospectively for a total of 1916 days between 01 January 2009 and 31 March 2014. Response variables were number of the admission, hospitalization, and mortality due to asthma, COPD, pneumonia and pulmonary thromboembolism. The response variables were estimated by GAM, DGDOM and GAMAR methods by building 4 different models and the performances of the models were compared. Results: GAM and DGDOM showed similar results according to the performance criteria in predicting the recourse and hospitalization response variables of asthma, COPD, pneumonia and PTE diseases. It was observed that GAMAR method gave superior results to other methods. In the mortality response variables of the diseases; although DGDOM gave superior performance to other models in asthma disease, GAMAR was also superior in other diseases. Conclusion: When the models created with dependent variables was compared, it was observed that the best performance was obtained from the GAMAR method. However, the lowest standard error graphics in the OR charts were observed in the models using the GAMAR method. Key words: GAM, DGDOM, GAMAR, Environmental Epidemiology, Air Pollution.