Abstract:
The objective of this research is to study efficiently criterion of model selection and reliability of search for the best suitably correctness of model within the same data. The Mean Absolute Percentage Error (MAPE), which the statistic tool used, act as the criteria in this research. This research used 3 and 4 independent variables. The distribution of error is normal distribution with mean equal to 0 and standard deviation equal to 1, 2, 3 and 5, respectively. The sample sizes are 20, 35 and 50, respectively. The levels of correlation among the independent variables could be classified into 3 levels for which low levels equal to 0.05 - 0.35, middle levels equal to 0.40 - 0.65 and high levels equal to 0.70 - 0.95. The study used the Monte Carlo Simulation method. The experiment was replicated 500 times under each situation. The analyzed results of the data were demonstrated as follow. CASE 1 The number of independent variables is 31. a sample size = 20 a) standard deviation = 1, 2 and independent variables when they have correlation in middle levels of the selected model is reduced model which was subtracted one variable but if independent variables have correlation in high levels the selected model is reduced model that was subtracted two variables b) standard deviation = 3,5 and also independent variables when they have low levels of correlation and upper the selected model is reduced model that was subtracted one variable but if independent variables have correlation in high levels the selected model is reduced model that was subtracted two variables 3.a sample size = 35 a) standard deviation = 1, 2 and independent variables when they have correlation in middle levels and upper the selected model is reduced model which was subtracted one variable b) standard deviation = 3,5 and independent variables when they have correlation in middle levels and upper the selected model is reduced model which was subtracted one variable but if independent variables have correlation in high levels of the selected model is reduced model that was subtracted two variables 1. a sample size = 50 independent variables have correlation in high levels of the selected model is reduced model that was subtracted one variable In the other cases the selected model is a full model. CASE 2 The number of independent variables is 4 2. a sample size = 20 a) standard deviation = 1, 2 the result is the same as case 1 ( n = 20) b) standard deviation = 3, 5 and independent variables when they have correlation in low levels of the selected model is reduced model that was subtracted one variable but if independent variables have correlation in middle levels of the selected model is reduced model that was subtracted two variables and if independent variables have correlation in high levels of the selected model is reduced model that was subtracted three variables 3. a sample size = 35 the result is the same as case1 ( n = 35) 4. a sample size = 50 a) standard deviation = 1, 2 and independent variables when they have correlation in high levels of the selected model is reduced model that was subtracted one variable b) standard deviation = 3, 5 and independent variables when they have correlation in middle levels and upper the selected model is reduced model that was subtracted one variable In the other cases the selected model is a full model. The MAPE varies with level of correlations, standard deviation, number of independent variables but converses to sample sizes.