[argyllcms] Re: OT: some metrological questions
- From: Roger Breton <graxx@xxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 25 Jul 2008 11:49:12 -0400
Hallo Klaus,
Personally, and naively perhaps, I would stick with Lab over spkt
measurements. Perhaps Robin Myers (www.rmimaging.com) has another opinion?
I believe the average remains the best statistics but others may argue that
the normality assumption must be verified -- ask Gernot for help. I don't
know which tests these involve but I believe it's possible to test for
normal distribution. Now, these tests involve more than 5 or 10
occurances...
Beyond that, I understand that you'd need some kind of weighting method for
computing the average. Similar to what MeasureTool uses.
For characterizing the deviation, I think the best statistics remain
standard deviation. But, again, I suspect you need at least n=30
measurements to make meaningful statistical inferences from the data.
MfG,
> When I have spectral or tristimulus measurement data with repeated
> patches (e.g. 5x white, 3x red, 2x green, 10x blue, ...): what is the
> best method to average them?
>
[snip...]
> - Working directly with the spectral data is probably the most
> appropriate (but expensive)-: method, but I have no idea e.g how take
> the correlations between the bands into account.
>
> Thanks in advance for any help,
> Klaus
Roger Breton
Other related posts:
