Most stringent test of independence for time series

Abstract

Evaluating the relationship between two variables has been and will remain the core objective of practitioners in social sciences especially in economics. After the criticism of the use of Karl Pearson Product Moment Coefficient of Correlation (PPMCC) by Yule (1926), several tests of independence for time series have been developed. This study aims at conducting a comparative study of these tests of independence for time series by using a general data generating process. We have made use of a Monte Carlo simulation study which compares these tests through the standard stringency criterion as discussed in Asad Zaman (1996). Among eleven tests of independence, two tests have been found to be in the best category, one test in mediocre and the remaining eight tests in worst category according to stringency criterion. Results have shown that the tests based on the idea of pre-whitening the series’ first and then to apply correlation coefficient have produced better results.

Keywords:

• Comparison
• Monte Carlo Simulation
• Pre-whitening
• Size and power
• Stringency
• Tests of independence
• Time series