Re: Efficiency Tests?

[TomDoan]

Campbell, Lo and MacKinlay has a rather long chapter on tests of the RW hypothesis, starting with the most restrictive interpretation (i.i.d. mean zero increments), loosening it to simply independent, and finally to uncorrelated. You might want to take a look at that. (Someone else might be able to suggest a more recent reference). If you want to do a run test, you can use the @RUNTEST procedure. You can also check out the BDINDTESTS procedure, which does a whole set of non-parametric tests for independence.

[menykeko]

Dear Tom,
I downloaded the Runtest procedure but am finding it difficult to estimate it. Each time I try, I get the following result:
@RUNTEST ET
Run Test for Series ET
1068 Runs in 1068 Observations. Prob(success) = 0.00000
Z-score NA p-value NA

How do I get it right?

Thanks in advance!

[TomDoan]

Note the second line in the procedure: computes a run test for an input series which shows two states (1,0). you have to define the dummy variable series that represents the two states and input that into the procedure. I just posted a new version which includes a validation check for the data at @RUNTEST

[menykeko]

Dear Tom,
I checked out the BDINDTESTS as you suggested above for nonparametric independence tests and obtained estimates for rank test, difference sign, and turning point.
But I could not find their any other information, such as how to specify them, in any of the RATS manual.
Do you have any solution regarding specifying them or any material that I can look at?

Thanks

[TomDoan]

They're obviously defined in Brockwell and Davis. I assume you can find descriptions of any of them on Wikipedia.

[menykeko]

I have found them.
Thanks