Iis Dewi Ratih, Sutikno, Purhadi : Copula Regression to Modeling The Rice Harvest in Jember Regency

Iis Dewi Ratih S.Si., M.Si.
Dr. Sutikno S.Si, M.Si.
Drs. Purhadi M.Sc



Published in

Annual Basic Science International Conference

External link


Seminar Internasional


Regression, Copula, Rainfall, Rice Harvest Area


The relationship between some of random variable is one of issue that is very much studied in statistical sciences. Some dependencies methods such as Pearson correlation and regression OLS still requires assumptions that rarely met in the real case application. Copula is a statistical method which has many advantages in modeling dependencies between variables. Some advantages of copula are invariant to the transformation, not strict on the distribution assumption, could explain he nonlinear dependencies, and easily construct the joint distribusion function. Copula regression is a regression-based Copula that describes causal relationships. Regression copula not have regression parameters as in the other regression models. Copula regression only have copula parameters and marginal function parameters for each random variable marginal. The method used to estimate the parameter is Maximum Likelihood Estimation (MLE) and Copula used in regression models only Copula gaussian. Performance of copula will be applied, to the modeling of rice harvested area (RH). RH consists of three periods, namely the rice harvested area 1 (RH1), rice harvested area 2 (RH2) and the rice harvested area 3 (RH3) with independent variable is rainfall in the largest rice production centers, Jember. The results showed that copula regression provides better than OLS (Ordinary Least Square), GLM (Generalized Linear Models), and GCMR (Gaussian Copula Marginal Regression) to modeling RH2, because gives the smallest RMSE. While RMSE Copula regression for RH1 and RH3 has a small difference value with GCMR, because Copula used in the regression modeling is Copula gaussian. Actually, not all the variables of rainfall have a linear relationship (follow Copula gaussian) with variable rice harvested area. So that, for modeling the rice harvested area are advised to use Archimedean Copula in copula regression.