"Tom Lane" <tlane@mathworks.com> wrote in message <l3n2m0$pj9$1@newscl01ah.mathworks.com>...
> > I have 6 images that look almost the same, there are some differences
> > though. I've withdrawn 5 of images from the first one and I'm trying to
> > use anova1 to see if they are from the same population but I get strange
> > results. Figure 2 shows red '+' signs (the actual values of my data) and
> > it is from 0 to 255. And there is a blue line at 0. What am I doing wrong?
>
> I don't understand exactly what you are doing. However, here is something
> that will reproduce what you saw:
>
> x = zeros(1000,5);
> x(ceil(5000*rand(100,1))) = floor(255*rand(100,1));
> anova1(x)
>
> Now, I understand this is not what you have done. But I have data mostly at
> or near zero, with a relatively small proportion of points taking other
> values as high as 255. This may describe what your image matrices look like.
> If you make a box plot of them (your Figure 2), you should see that the
> quartiles and medians are drawn as blue lines right near zero, and the other
> values are outliers drawn as red plus signs above that.
>
> The other figure will contain your anova table. Normally anova is performed
> on continuous data. You appear to have coarse data with lots of zero values.
> I don't know if anova would work well for that.
>
> I'd have to understand more about your problem to suggest anything.
>
>  Tom
Thank you for the answer, Tom!
Yes, I get that kind of box plot. Okay, I can try and explain my problem. I have 6 images, one is original and 5 are slightly changed in a comparison to the original image. I've subtracted the 5 changed images from the original one and got 5 populations. And I'd like to know if these 5 slightly changed images (the methods that made them different) comes from the populations with the same mean...
