Q-Q Plot

Damon_McDougall2 · August 15, 2012, 11:02pm

Hi all,

If anyone hasn't noticed already. I've been somewhat semi-perusing the
Matlab interface and trying to port over any plotting functionality not
current present in mpl.

The other day I made a bit of a mess up regarding functionality I didn't
think existed but actually did (discussion here:
https://github.com/matplotlib/matplotlib/pull/1068)

I'm sending a message here because I've found another one I don't think
currently exists in mpl:
http://www.mathworks.co.uk/help/toolbox/stats/qqplot.html

I wanted to check with everyone here (who probably know better!) before
I started anything.

Thanks!
Damon

···

--
Damon McDougall
http://www.damon-is-a-geek.com
B2.39
Mathematics Institute
University of Warwick
Coventry
West Midlands
CV4 7AL
United Kingdom

_Paul_Hobson · August 15, 2012, 11:56pm

Damon,

When the current state of the relevant python libraries, scipy is
required to create a QQ plot. Therefore, matplotib will never be able
to natively make QQ or probability plots. I've got a PR into the the
statsmodels project to do just what you need (and more!).

github.com/statsmodels/statsmodels

Q-Q plot with probabilities

statsmodels:master ← phobson:qqplot-with-probs

opened 02:07PM - 31 Jul 12 UTC

phobson

+619 -58

In environmental engineering, bosses and regulators prefer to see probability pl…ots over quantile plots since you can more easily pick out the likelihood of meeting (or exceeding) regulatory limit. It was pretty simple to add this option to the qqplot function, so I simply expanded the existing test to cover it. Here's an example plot: ``` python import numpy as np import statsmodels.api as sm import matplotlib.pyplot as plt x = np.random.normal(4.0, 1.75, size=37) fig, (ax1, ax2) = plt.subplots(nrows=2) sm.qqplot(x, prob=False, ax=ax1) sm.qqplot(x, prob=True, ax=ax2) ``` ![QQPlot with probabilities](https://lh6.googleusercontent.com/-cRNHCw-PaGQ/UBfiCfypYjI/AAAAAAAABDo/S03wPcviLmU/s800/qqtest.png) And a public gist: https://gist.github.com/3217134 At the moment, I'm just testing the waters with this PR. Ideally, I'd like to have 1 more option for the `prob` keyword beyond just `True` or `False`. Perhaps, `None,`'under', 'over'`. - `None` would show the theoretical quantiles on the x-axis (existing functionality, top axis). - `'under'` would create example like the link above (currently in this PR and lower axis). - `'over'`, would flip the quantiles with `theoretical_quantiles` *= -1, making the x-axis display the likelihood of the sample set exceeding a given value (proposed). There's certainly better choices for those options, but they escape me at the moment. Thanks for any feedback and taking the time to look at this.

It'll probably be a while before I get that PR up to par with the rest
of statsmodels, so in the meantime, I suggest you use the existing
functionality in statsmodels or roll your own.

Good luck,
-paul

···

On Wed, Aug 15, 2012 at 4:02 PM, Damon McDougall <damon.mcdougall@...149...> wrote:

Hi all,

If anyone hasn't noticed already. I've been somewhat semi-perusing the
Matlab interface and trying to port over any plotting functionality not
current present in mpl.

The other day I made a bit of a mess up regarding functionality I didn't
think existed but actually did (discussion here:
Add stairstep plotting functionality by dmcdougall · Pull Request #1068 · matplotlib/matplotlib · GitHub)

I'm sending a message here because I've found another one I don't think
currently exists in mpl:
Quantile-quantile plot - MATLAB qqplot - MathWorks United Kingdom

I wanted to check with everyone here (who probably know better!) before
I started anything.

Thanks!
Damon

--
Damon McDougall
http://www.damon-is-a-geek.com
B2.39
Mathematics Institute
University of Warwick
Coventry
West Midlands
CV4 7AL
United Kingdom

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
matplotlib-devel List Signup and Options