Testing for Heteroscedasticity in Multiple-Input-Single-Output (MISO) System Regression Model

In this study of an input/output system, often using the regression model to forecast and analysis. In the classical regression analysis, usually assumes that the system has the observation value of homogeneity of variance, it can make the routine statistical inference. If the variance of the system of observations is non homogeneous, then the regression analysis will be difficult to accurately. In practice, the system of observation value is difficult to determine its homogeneity of variance. Therefore, to predict when the regression model, the homogeneity of variance testis very necessary. The main work in this paper is the likelihood ratio test statistic and Score test statistic comparison of multi input single output system model, and comparing the likelihood ratio test statistic and Score test statistic modified orthogonalization parameters obtained. This paper use Monte-Carlo method to simulate the effect of several kinds of Statistics. Simulation results show that: the effect of several test statistics have good test effect.