quantreg: exception with singular exog

I’m trying to run quantile regression with a number of variables. It appears quantreg.fit() will not allow more than 8 variables/columns to be included in the formula.

import pandas as pd 
import statsmodels.formula.api as smf
from statsmodels.regression.quantile_regression import QuantReg

d = pd.read_csv('<path-to-data>')

#7 vars (lag and sunday through friday), seems to work fine
qr = smf.quantreg('load ~ lag + su + m + t + w + th + f',  d')
qm = qr.fit(q=0.5)

#8 vars (lag and sunday through saturday), throws error
qr = smf.quantreg('load ~ lag + su + m + t + w + th + f + s',  d')
qm = qr.fit(q=0.5)

Here’s the trace:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-17-f2074b49225f> in <module>()
---> 32 qm = qr.fit(q=.5)

.../statsmodels/regression/quantile_regression.py in fit(self, q, vcov, kernel, bandwidth, max_iter, p_tol, **kwargs)
    177             resid = np.abs(resid)
    178             xstar = exog / resid[:, np.newaxis]
--> 179             diff = np.max(np.abs(beta - beta0))
    180             history['params'].append(beta)
    181             history['mse'].append(np.mean(resid*resid))

ValueError: operands could not be broadcast together with shapes (9,) (8,) 

[hpahkala]

I second more informative warning or message. I had collinearity problem in my data but started to debug the issue from the wrong place due to the message.

[josef-pkt]

fixed in #7694 based on PR #6397

The behavior now is that it doesn’t raise on singular exog, but instead provides a pinv solution.