On Wed, 27 Jun 2007, Malcolm J. Currie wrote:
> Tim,
>
>> :-) I did read the documentation. There was no reference as to where the
>> magic number came from.
>
> See Parameter MODE.
>
which doesn't say where the estimate comes from.
>> Example data file attached. Does a nice job in histogram with 64 bins.
>
> A concern is coping with the general dataset rather than a specific
> one or class.
>
> One could make an intelligent guess for the number of bins based
> upon the data range or clipped standard deviation, and the number of
> values. This could be a dynamic default.
ok
>
>> The mean is miles out and the median is slightly lower than the peak.
>
> Do you mean (pun unintended) the results from HISTAT applied to the
> example file?
Yes. The answers are correct, I was simply justifying my need for MODE in
this data set.
>
>> sounds good. Does it really need multiple MODEs? If the user is doing this
>
> For a coarse bin you're unlikely to be multi-modal although there may be
> multiple peaks in the histogram. I was thinking more of the narrow bins
> needed to derive sufficient accuracy for percentiles.
>
>> are they allowed to take the risk that it picks the wrong one? I suppose
>> it could return * if the peaks are about the same height.
>
> Another precise specification. (-:
>
>> Didn't know about HISTPEAK.
>
> I'm surprised that it's not in the HISTAT "Related Applications".
>
Phew. I thought that you were upbraiding me for not reading the "Related
Applications" section (Which I hadn't done) but I just looked so I'm in
the clear.
>> Tried it and it moaned at me saying that the
>> data didn't have enough spread. Presumably it chose too many bins.
>
> I've used HISTPEAK in my research. Given a skewed unimodal distribution
> of a sky plus noise and sources that it expects, HISTPEAK used to work
> quite well.
>
> The lack of spread rings bells. ESP assumes counts from some CCD or
> photographic scan and might not like very small numbers.
>
>> I'm happy for you to decide. If HISTPEAK can be made to work then that
>> would be fine. Is it a lot of work to tweak HISTAT or are there kaplibs
>> routines that can be used?
>
> Are you suggesting a free transfer to KAPPA? Not sure HISTPEAK would
> pass the medical and fitness tests.
>
Fixing HISTPEAK in ESP would be fine.
> You'd think it would be less work to hack HISTPEAK, but my money is on
> HISTAT if I just find the peak of a coarser histogram but not fitting.
>
Fine with me. Your choice.
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|