Re: hist.m differences from Matlab

[Doug Stewart]

On Wed, Dec 1, 2010 at 1:02 PM, Rik <address@hidden> wrote:
>>> So for non uniform bin spacing
>>> - find max and min
>>> - find point half way between each given point
>>> - construct a group of ranges from the above
>>> - count how many are in each of these non uniform ranges.
>>> Does this look like something we should implement?
>>> If yes I will work on it latter.
>>> Doug
>>
>> Yes, I think this should behave as ML does.
> I believe Octave already does all of this correctly.  If I run
> [a,b] = hist(y,x)
>
> a =
>
>   562    78    91    55   214
>
> b =
>
>   0.10000   0.20000   0.50000   0.70000   0.90000
>
> which shows that it has the bin centers correct and the counts are mostly
> correct as well.  Since randn is normally distributed I would expect about
> half the results to be less than zero or about 500.  Because the first bin
> center is at 0.1 rather than zero the count is slightly larger than 500.
>
> It seems rather that the bar plot does not reproduce the Matlab plot
> because Octave uses fixed width bars for the bins while Matlab uses bars
> that correspond to the non-uniform bin width.
>
> --Rik
>

Good point Rik
I was out all afternoon so haven't looked at it yet.
Doug