I've noticed an oddity with histat when called with the percentiles
argument. I want to know the value for a given percentile so I pass in
a single value (say, percentiles=75).
Under some conditions (which I haven't been able to identify) when I
try and retrieve the PERVAL parameter I find it's an array with 3
elements, the first two (or sometimes last two) of which are bad.
I can reproduce this in ORAC-DR when processing multiple files. The
first file is fine - there is only a single value for perval - but the
second file shows perval is an array as above. (Trying it again with
multiple percentiles shows again that the first two elements are bad.)
Ah, it may be related to the mode calculation. With method=pearson it
doesn't seem to happen. I can reproduce it on the command line if I
specify method=moments or method=histogram.
Example file here, example commands below:
http://www.astro.ubc.ca/~agg/starlink/s8d20100307_00012_0001_noi.sdf
% histat s8d20100307_00012_0001_noi.more.smurf.nep percentiles=75 method=moments
% hdstrace ~/adam/histat.sdf
<stuff deleted>
PERVAL(3) <_DOUBLE> *,*,1.23914414099084D-16
% histat s8d20100307_00012_0001_noi.more.smurf.nep percentiles=75 method=pearson
% hdstrace ~/adam/histat.sdf
<stuff deleted>
PERVAL(1) <_DOUBLE> 1.23914414099084D-16
Has anyone else noticed this?
Andy
|