It’s awesome to find sampling distribution of Durban-Watson test statistic that is not reveal any sampling distribution before. So it has been very doubt to me for a long time.
Yes. Many factors affect Durban-Watson test statistic simultaneously. That’s why the inconclusive exists. The website shows how factors affect sampling distribution of Durban-Watson test statistic, respectively.
But the outcomes of that website are from a software for Durban-Watson test. It’s amazing that the developers release the software to be used by students, scholars, researchers, and so on.
They also offer the source code of the software including how to run five multithreading.
Back to the software, it is based on the simulator of probability distribution. Then the software can get the sampling distribution by running 100 million times to generate T samples per time.
Another great function is that the user can set parameters by himself or herself. Parameters include sample size number (T), autocorrelation coefficient (tho), lags of errors. They provide four models to satisfy the needs of the users. Model II has to add additional parameter of Y bar(sample mean of Y). Model III and Model IV need the values of independent variable(s). There is a folder, named “independent”, to save the values of independent variable(s).
This is a flexible software for users to run sampling distribution and get the critical values. Users can try any setting of parameters and then compare the sampling distribution or test the serial correlation. They will not be limited in the test of whether autocorrelation is zero or not, but try the test for any value of autocorrelation coefficient.
Download website: https://tinyurl.com/dwtestsoftware.
Video to display features: https://youtu.be/BLGMxQC8a3M
Big Data analysis and simulator 