Abstract:
The objective of this study is to find out the cut-off point estimation of binary logistic regression model for predictive classification using complementary log-log function as a link function. The interesting factors are failure rate (a) of the values 0.25,0.5, and 0.75, number of independent variables (p) are 1, 2, 3, 4, and 5, sample size (n) are 50,100,150,200 and 250. Degrees of multicollinearity among independent variables with 3 levels are low, medium, and high level. The data are generated using Monte Carlo technique through R-program. The results are summarized as follow. As the failure rates change while keeping other factors constant, only one number of independent variable gives the mean value of the cut-off points not exceeding 0.5 in the switching manner. When the number of independent variables increase to 2, 3, 4, and 5, the mean value of the cut-off points increases as well but none of them exceed 0.5. As the sample size changes and the other factors are kept constant, the number of independent variables of 1 and 2 gives the mean value of the cut-off points not exceeding 0.5; also switching. When the number of independent variables increases to 3, 4, and 5, most of the mean values of the cut-off points increase but not exceeding the value of 0.5. As the number of the independent variables change and the other factors are kept constant, the mean value of the cut-off points is found to be increased at every level of the sample sizes. Finally, as the failure rates change when the other factors are kept constant; at the number of independent variable equal to 1, the mean value of the cut-off points is found to be constant not exceeding 0.5. When the number of independent variables increase to 2, 3, 4, and 5, the mean value of the cut-off points increase.