ESTIMASI PARAMETER MODEL REGRESI LOGISTIK MENGGUNAKAN METODE RESIDUAL BOOTSTRAP; ESTIMATION OF LOGISTIC REGRESSION MODEL PARAMETER USING BOOTSTRAP RESIDUAL METHOD

FATIMAH, SYAHESTI NURUL, Herni Utami

2016 | Skripsi | FMIPA

Bootstrap is one of the estimation methods, computer-based statistical inference. Its working principle is using a computer in generating original data from a small sample to get a pseudo sample. Pseudo sample is obtained by taking samples random with repetition from the original sample can then be used to calculate the value of the estimator. This pseudo sample took with a replacement from the original sample. There are three Bootstrap procedures to obtain pseudo sample, namely the Bootstrap based on residuals, the paired Bootstrap, and the external Bootstrap. The Bootstrap based on residuals is better method because it will produce a small standar error values. The Bootstrap method?s advantage is confidence interval percentiles Bootstrap length shorter than the others. The main purpose of this method is to obtain the best possible estimate based on minimal data with the help of computers. In this paper, a Bootstrap method is applied to estimate the parameters of a logistic regression model. Logistic regression model is a form of regression analysis to determine a causal relationship (causality) when the response variable Y has only two possible values or data are dichotomous. The method which is often used to solve the logistic regression problem is Maximum Likelihood Estimation (MLE) where the parameter estimation process is preceded by the formation of likelihood function. Bootstrap method in estimating parameters of the logistic regression model is applied in the determination of how much influence factors of hypercholesterolemia from sample random patients Yogyakarta Health Laboratory. Based on the results of the analysis, Bootstrap method is able to reduce the standar errors to Bootstrap with repetition as much as 1000 times.

Kata Kunci : Bootstrap Method, Logistic Regression, Maximum Likelihood Estimation (MLE)