%0 Journal Article
%T MULTIPLE LOGISTIC REGRESSION MODEL TO PREDICT RISK FACTORS OF ORAL HEALTH DISEASES
%A Shivalingappa B. Javali
%A Parameshwar V. Pandit
%J Revista Roman£¿ de Statistic£¿
%D 2012
%I Romanian National Institute of Statistics
%X Purpose: To analysis the dependence of oral health diseases i.e. dental caries and periodontal disease on considering the number of risk factors through the applications of logistic regression model. Method: The cross sectional study involves a systematic random sample of 1760 permanent dentition aged between 18-40 years in Dharwad, Karnataka, India. Dharwad is situated in North Karnataka. The mean age was 34.26¡À7.28. The risk factors of dental caries and periodontal disease were established by multiple logistic regression model using SPSS statistical software. Results: The factors like frequency of brushing, timings of cleaning teeth and type of toothpastes are significant persistent predictors of dental caries and periodontal disease. The log likelihood value of full model is ¨C1013.1364 and Akaike¡¯s Information Criterion (AIC) is 1.1752 as compared to reduced regression model are -1019.8106 and 1.1748 respectively for dental caries. But, the log likelihood value of full model is ¨C1085.7876 and AIC is 1.2577 followed by reduced regression model are -1019.8106 and 1.1748 respectively for periodontal disease. The area under Receiver Operating Characteristic (ROC) curve for the dental caries is 0.7509 (full model) and 0.7447 (reduced model); the ROC for the periodontal disease is 0.6128 (full model) and 0.5821 (reduced model). Conclusions: The frequency of brushing, timings of cleaning teeth and type of toothpastes are main signifi cant risk factors of dental caries and periodontal disease. The fitting performance of reduced logistic regression model is slightly a better fit as compared to full logistic regression model in identifying the these risk factors for both dichotomous dental caries and periodontal disease.
%K Dental caries
%K Periodontal disease
%K Akaike Information Criterion
%K Receiver Operating Characteristic
%K Full model
%K Reduced model
%U http://www.revistadestatistica.ro/Articole/2012/art7en_rrs_5_2012.pdf