KLASIFIKASI SKOR PROPENSITAS DALAM PENDUGAAN SELANG KEPERCAYAAN BOOTSTRAP UNTUK PERBEDAAN NILAI TENGAH DUA POPULASI
The comparison of mean of two populations assumes that there is no other variables influence (covariate) except the difference of the observed variable. In real data, this condition is often unfulfilled. Propensity score classification (PSC) is a method to overcome the case. In this research, we do simulation data to evaluate the method, and as the illustration we do the real data of first semester NMR of IPB postgraduate students in Statistics major. The simulation data is generated by covariates either with the same means or different ones to both groups, each with different parameter () 0,00, 0,25 and 0,50. The bootstrap confidence interval included a distribution which is built from propensity score estimations without the variance estimation.
The result shows that 95% bootstrap confidence interval with PSC method includes the parameter for the same and different covariate distribution respectively as 0,95 and 0,87. This method is suitable only when the sample sizes are larger. The illustration uses real data with covariate age, marital status, graduate (S-1) NMR and occupation as a lecturer or not, the result estimation of 95% bootstrap interval confidence to differentiate NMR of postgraduate Statistics students in IPB between those who came from the universities in Java and outside Java is between -0,13 and 0,72.
Keywords : propensity, bootstrap, covariates