Error after update

[Nick Alexander]

Hi Daniel,

I updated copulae to the newest version 0.7.3 and the notebook I previously used to test copulae no longer works. Here is a snapshot of the first part of the code:

import copulae
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import scipy.stats as st

from statsmodels.distributions.empirical_distribution import ECDF
from copulae import ClaytonCopula, FrankCopula, GumbelCopula

np.random.seed(100)

v1 = np.sort(st.norm.rvs(loc=0,scale=1.5,size=10000))
v2 = st.norm.rvs(loc=0,scale=1,size=10000)

v3 = np.sort(st.norm.rvs(loc=0,scale=1.5,size=10000))
v4 = st.norm.rvs(loc=0,scale=1,size=10000)

total_sims = pd.DataFrame([0.1*(v1+v2),0.1*(v3+v4)],index=['V1','V2']).T

plt.scatter(total_sims['V1'], total_sims['V2']);

clay_cop = ClaytonCopula(theta=1,dim=2)
clay_cop.fit(data=total_sims, x0=None)

This code yields the following error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-1-c5dff29e8c64> in <module>
     21 
     22 clay_cop = ClaytonCopula(theta=1,dim=2)
---> 23 clay_cop.fit(data=total_sims, x0=None)

~\Miniconda3\lib\site-packages\copulae\copula\base.py in fit(self, data, x0, method, optim_options, ties, verbose, to_pobs, **kwargs)
    144 
    145         x0 = np.asarray(x0) if x0 is not None and not isinstance(x0, np.ndarray) and isinstance(x0, Collection) else x0
--> 146         self._fit_smry = fit_copula(self, data, x0, method, verbose, optim_options, kwargs.get('scale', 1))
    147 
    148         if isinstance(data, pd.DataFrame):

~\Miniconda3\lib\site-packages\copulae\copula\estimator\estimator.py in fit_copula(copula, data, x0, method, verbose, optim_options, scale)
     92     m = method.lower()
     93     if m in {'ml'}:
---> 94         x0 = initial_params(copula, data, x0)
     95         return estimate_max_likelihood_params(copula, data, x0, options, verbose, scale)
     96     elif m in ('itau', 'irho'):

~\Miniconda3\lib\site-packages\copulae\copula\estimator\estimator.py in initial_params(copula, data, x0)
    117 
    118     try:
--> 119         start = estimate_corr_inverse_params(copula, data, 'itau').params
    120         ll = copula.log_lik(data)
    121 

~\Miniconda3\lib\site-packages\copulae\copula\estimator\corr_inversion.py in estimate_corr_inverse_params(copula, data, type_)
     32         raise ValueError("Correlation Inversion must be either 'itau' or 'irho'")
     33 
---> 34     icor = fit_cor(copula, data, type_)
     35 
     36     if is_elliptical(copula):

~\Miniconda3\lib\site-packages\copulae\copula\estimator\corr_inversion.py in fit_cor(copula, data, typ)
     69     indices = tri_indices(copula.dim, 1, 'lower')
     70     if typ == 'itau':
---> 71         tau = kendall_tau(data)[indices]
     72         theta = copula.itau(tau)
     73     elif typ == 'irho':

~\Miniconda3\lib\site-packages\copulae\stats\correlation.py in kendall_tau(x, y, use)
    162     %s
    163     """
--> 164     return corr(x, y, 'kendall', use)
    165 
    166 

~\Miniconda3\lib\site-packages\copulae\stats\correlation.py in corr(x, y, method, use)
    101             raise ValueError('x must be a matrix with dimension 2')
    102         c = np.identity(x.shape[1])
--> 103         for (i, j), (c1, c2) in _yield_vectors(x, use):
    104             c[i, j] = c[j, i] = compute_corr(c1, c2)
    105 

~\Miniconda3\lib\site-packages\copulae\stats\correlation.py in _yield_vectors(x, use)
    256                 yield (i, j), (x[v, i], x[v, j])
    257             else:
--> 258                 yield (i, j), (x[:, i], x[:, j])
    259 
    260 

~\Miniconda3\lib\site-packages\pandas\core\frame.py in __getitem__(self, key)
   2900             if self.columns.nlevels > 1:
   2901                 return self._getitem_multilevel(key)
-> 2902             indexer = self.columns.get_loc(key)
   2903             if is_integer(indexer):
   2904                 indexer = [indexer]

~\Miniconda3\lib\site-packages\pandas\core\indexes\base.py in get_loc(self, key, method, tolerance)
   2893             casted_key = self._maybe_cast_indexer(key)
   2894             try:
-> 2895                 return self._engine.get_loc(casted_key)
   2896             except KeyError as err:
   2897                 raise KeyError(key) from err

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()

TypeError: '(slice(None, None, None), 0)' is an invalid key

Given this error did not occur with the previous version, is it possible there is a bug in the update? I played around with the inputs to the fit function with no success. Thanks!

[DannieB]

Aye, replicated it too. Issue arose because I wanted to be more permissive of pandas dataframes. Didn't catch an edge case so I've added tests for it.

The fix should be pushed up by tomorrow

[DannieB]

Should be fixed now.