ESTIMASI PARAMETER MODEL REGRESI LOGISTIK BINER MENGGUNAKAN METODE JACKKNIFE

Abstract

Logistic regression is one of the statistical methods used to identify a causal relationship (causality) between one dependent variable and one or more independent variables, where the dependent variable has 2 or more categories. To find a good logistic regression model, sufficient amount of data is needed, if the amount of data is insufficient, the modeling can be done using the resampling method, namely using Bootstrap or Jackknife. The working principle of the Jackknife method use computers in generating data from the original samples to generate artificial samples. An artificial sample is generated by removing an observation from the original sample which can then be used to calculate estimator value. In this case, the Jackknife Method is used to resampling the data of beras miskin admission in Yogyakarta. Based on the results of the analysis, Jackknife method can minimize the standard error in calculating the estimation of parameters until Jackknife Method erases two datas. Based on data obtained from SUSENAS (Sensus Sosial Ekonomi Nasional) of Yogyakarta  in 2016, there are 641 households being sample. The dependent variable which is used was the admission of beras miskin. Beras miskin consists of two categories. That is receiving beras miskin and not-receiving beras miskin. From the number of households affected by sample, it shows that the number of households receiving beras miskin is 15% (98 households) and the number of households which isn’t receiving beras miskin is  85% (543 households). The significant variables after logistic regression test were the main floor of house (X₂) and the source of drinking water (X4). Those variables affect admission of beras miskin with the odds ratio =0,2476 and =0,1999 . The best logistic regression model for case studies discussed in the previous chapter is a logistic regression model with the  two datas erased by jackknife method, in which the probability model of a household receiving raskin is:

P(Y = 1|X) = π (x) =

With the level of accuracy of the model in predicting at 84.2%