Permutation Methods for Chow Test Analysis an Alternative for Detecting Structural Break in Linear Models

摘要

This study examined the performance of two proposed permutation methods for Chow test analysis and the Milek permutation method for testing structural break in linear models. The proposed permutation methods are: (1) permute object of dependent variable and (2) permute object of the predicted dependent variable. Simulation from gamma distribution and standard normal distribution were used to evaluate the performance of the methods. Also, secondary data were used to illustrate a real-life application of the methods. The findings of the study showed that method 1(permute object of dependent variable) and Method2 (permute object of the predicted dependent variable) performed better than the traditional Chow test analysis while the Chow test analysis was found to perform better than the Milek permutation for structural break. The methods were used to test whether the introduction of Nigeria Electricity Regulatory Commission (NERC) in the year 2005 has significant impact on economic growth in Nigeria. The result revealed that all the methods were able to detect presence of structural break at break point 2005. Also, the methods were used to test for structural break at January, 2015 for monthly reported cases of appendicitis in Nigeria. Result revealed that all the methods were able to detect presence of structural break at break point January, 2015.