AMIN TOHARI, NUR CHAMIDAH* AND FATMAWATI
Department of Mathematics, Faculty of Science and Technology, Universitas Airlangga, Indonesia
*(e-mail : nur-c@fst.unair.ac.id; Phone: +62315936501)
ABSTRACT
As in June 2018, 433 (84.2%) out of 514 districts/cities in 34 provinces in Indonesia reported HIV/AIDS. Therefore, goal of this study was to model the number of HIV and AIDS cases in Indonesia by using biresponse Negative Binomial non-parametric regression based on the local linear estimator. The best model of HIV and AIDS cases with optimal bandwidth of 0.101 and deviance value of 37.73 was obtained. This was smaller than the deviance value of the parametric regression approach i. e. 47.36. It meant that biresponse non-parametric regression approach based on the local linear estimator was better than the parametric regression approach. For example, in Maluku province, if there was an increase in 1% of drug users, the number of HIV cases will increase by 1.51 times than before, and the number of AIDS cases will increase by 1.37 times than before.
Key words : HIV, AIDS, Negative Binomial regression approach, local linear estimator