boxplot: specifying data distribution ?

[CdeMills]

Hello,

I use boxplot on segmented data to detect portions where the acquisition
process went wrong. The data are the produced by injecting a pseudo-random
sequence into a system and recording its output.

For each segment, I extract a mean and a variance, collate everything and
group according to the input level values. Boxplot on the mean values works
fine. But on the variances, the undelying distribution is Fisher instead of
normal. The implication is that some of the points are flagged as outliers,
yet according to the Fisher pdf there are acceptable.

Would it be possible to specify the pdf of the data for boxplot ? 

Regards

Pascal 

--
View this message in context: 
http://octave.1599824.n4.nabble.com/boxplot-specifying-data-distribution-tp4444135p4444135.html
Sent from the Octave - General mailing list archive at Nabble.com.