Since no reply has been posted yet I will give it a shot. runs.test uses the
normal approximation and in your case it returned a z score of -1.8732. This
z score has a cumulative probability of
pnorm(-1.8732,0,1)
[1] 0.03052039
If you are concerned about having too many runs and too few runs you would
select the "two.sided" option for runs.test, which gives a p-value of 0.0610
(0.0305 in each tail of the normal distribution). If you are concerned only
with too few runs you would select the "less" option, which will give a
p-value of 0.0305. Finally, if you are concerned only with too many runs you
would select the "greater" option which will give a p-value of 1-0.0305 =
0.9693. If your significance level is 0.05, you would compare 0.05 to 0.0610
and not reject the null hypothesis for the two-sided case and compare 0.05
to 0.0305 in the one-sided case and reject the null hypothesis. Note that
the normal approximation is OK for large samples but may give unacceptable
results for small samples. I am unaware of any packages in R that perform an
exact runs test.
Tom
>I have used "runs.test" (Package tseries) for computes the runs test
>for randomness , but I get this result:
>
>Runs test
>-1.8732 P-value = 0.0610
>
>Alternative Hypothesis : Two sided
>
>How can I interpret this result ?
[[alternative HTML version deleted]]
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
