[R] Permutation test of marked point pattern
Patrick Burns
pburns at pburns.seanet.com
Fri May 5 10:44:25 CEST 2006
For completeness of the archives, this problem seems
to have been solved via a private conversation.
From my perspective the problem was trying to complicate
a situation that was essentially solved. Instead of
trying to comprehend how to conform the present
situation to someone's implementation of a permutation
test, it is much easier to just do the computation.
Something like:
permvec <- numeric(1000)
for(i in 1:1000) permvec[i] <- statistic.from.permuted.data
p.value <- mean(permvec <= original.statistic)
I think the same is true of bootstrapping. This is more
controversial since there are some techniques -- such as
bias correction -- with bootstrapping that can improve
the results. However, virtually always in the work that I
do a naive bootstrap is good enough.
Patrick Burns
patrick at burns-stat.com
+44 (0)20 8525 0696
http://www.burns-stat.com
(home of S Poetry and "A Guide for the Unwilling S User")
Jean-Francois Savard wrote:
>Dear R users,
>
>I am trying to perform a hypothesis test on a marked point pattern. I
>would like to calculate the mean of the absolute value of the
>difference of marks between nearest neigbours, randomize the marks
>among points, then calculate this mean again. Ideally, I would test
>whether random mean values smaller than the observed mean value occur
>less than 5% of the time. I suppose 1000 permutations would be a
>reasonable starting point (the ppp object has 27 points).
>
>so far, I've figured out how to:
>
>-create a marked ppp object: ms.ppp
>
>-calculate my test statistic:
>teststat <- mean(abs(markstat(ms.ppp, diff, N=2)))
>
>-randomly allocate marks to a point pattern:
>Y <- rlabel (ms.ppp, labels=ms.ppp$marks, permute=TRUE)
>
>I have looked at perm.test{exactRankTests} and perused the R help
>archive but haven't been able to find how to work the permutation test
>with a marked ppp object.
>
>I thank you in advance for any help,
>
>
>J-F Savard
>Doctoral Candidate
>University of Maryland
>
>______________________________________________
>R-help at 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
>
>
>
>
More information about the R-help
mailing list