“The exact method is impossible. Will be forced to use approximate method. ”in “regstats”function？

I know that my linear regression didn't pass the Durbin Watson test and then the Matlab informed me "The exact method is impossible. Will be forced to use approximate method." after I used the "regstats" function. But I don't know what kind of approximate method were used to rectify the autocorrelation problem of my linear regression?

my program"backtest" could be simplified into:

regstats(y,x)

Warning is as follows:

The exact method is impossible. Will be forced to use approximate method. 
> In pvaluedw (line 46)
  In dwtest (line 96)
  In regstats (line 364)
  In backtest (line 14)

Thanks in advance

[lilijiang]

How can I ignore this warning and use linear method that I initially set? Thanks a lot

[Walter Roberson]

This warning does not mean that 'linear' was not used: this warning just means that the Durbin Watson test was not able to perform an exact calculation and has fallen back to an approximation instead.

You can turn off the warning by using

warning('off', 'stats:pvaluedw:ExactUnavailable');

This will just prevent it from being displayed, but will otherwise not change anything.

Answer 1

[dpb]

See the Algorithms section under

doc dwtest

It is a normal approximation referenced back to the D-W article from Biometrika, lo! those many years ago.

[2 1]

Thanks a lot!