0

I run the durbin-watson test over my variables using 'dwtest' command. There are 8 independent variables and 267 samples. I get the following result and wonder whether I can conclude I have no autocorrelation problem. If not, how should I claim it?

      Durbin-Watson test                    

 data:  y ~ x + x2 + x3 + x4 + x5 + x6 + x7 + x8                    
 DW = 2.816, p-value = 1                    
 alternative hypothesis: true autocorrelation is greater than 0
Eric
  • 528
  • 1
  • 8
  • 26

1 Answers1

2

The p-value here is 1 (as big as you can get) which means you can't reject the null hypothesis (at any significance level.) The null hypothesis in this case is that the autocorrelation is zero.

Thus... There is not significant evidence that there is a non-zero autocorrelation.

Erik Westwig
  • 713
  • 4
  • 16
  • Thank you. I've seen there is a table for critical values for Durbin-Watson test. Based on this table my value is above the (upper) threshold. Is it ok to entirely trust the R outcome I have in this case? – Eric Aug 27 '17 at 12:46
  • Not sure exactly what you're asking... You likely can trust that the p-value calculation is correct if you see that your test value is greater than all the critical values. – Erik Westwig Aug 27 '17 at 13:18
  • The above link provides the critical values of Durbin-Watson test I am taking about. – Eric Aug 27 '17 at 17:26
  • Oh sorry I got it. The result I have is consistent with the manuscript I provided with the link. – Eric Aug 27 '17 at 18:31