Negatives values are ignored. Why?

Dear all, I am plotting few values that go from close to 0 to something large, so I want to the y axis to become in log scale.

boxplot(A','symbol','','extrememode','compress','labels',[1 2 3 4 5 6 7],'whisker',0); set(gca,'YScale','log') ; Warning: Negative data ignored

which I can understand as >> min(min(A))

ans =

    0.0015

>> log(min(min(A)))

ans =

   -6.4852

what do you think from your experience I should do for that? B.R Alex

Answer 1

[Daniel Shub]

If I understand boxplot correctly, it is not just plotting A, but it is also plotting errorbars and boxes. It is possible that the errorbars/boxes/whiskers become negative. I do not think this can happen with whisker 0 option. It might be that extrememode compress causes some of your data to become negative.

[Oleg Komarov]

h = boxplot(rand(100,1)) causes the same problem and it just how data is visualized.

get(h(4),'Ydata') will be positive but to draw properly the last whicsker the axis Y limit is set slightly negative.

[Alex]

I tried even this

ylim([-100 100]);

>> set(gca,'YScale','log');

Warning: Negative data ignored

which makes all the boxplots centered and pretty tine

I keep even getting the same warning...

[Oleg Komarov]

The natural log's domain is R^plus, i.e. > 0, and log(0) is -inf. You can't supply negative values to log.

http://en.wikipedia.org/wiki/Natural_logarithm