Allow regression metrics to work with significance=None

[jb-delafosse]

Reference Issue

See issue #6

What does this implement/fix? Explain your changes.

I added the default value None to various evaluation function for regression only

I think the same thing could be done for classification problems

[donlnz]

Thanks for the suggestion.

I agree that the evaluation metrics could support significance=None (thus evaluating validity/efficiency at all significance levels simultaneously). However, I find that there are some problems with your submitted patch: reg_mean_size, reg_median_size, etc should return (if significance=None), the mean, median (etc) size per significance level. Currently, they return the mean, median (etc) over all significance levels, which does not seem reasonable to me.

I also agree that a similar approach could be taken to classification evaluation metrics.

I'll see if I can get around to fixing these in the next few weeks. Otherwise, feel free to submit a new patch that solves the issues mentioned.

[jb-delafosse]

Ok. I see where I was wrong. I'm using only one significance level (95%) and hence it does what I want.

I need to make sure it works correctly when multiple significance levels are used.

Am I correct ?