Problem with example notebook American National Election Studies (ANES) data: ValueError: The first guess on the deviance function returned a nan. This could be a boundary problem and should be reported.

The example notebook with the election data is not working. The notebook in question is ANES_logistic_regression.ipynb.

import bambi as bmb
import pandas as pd
import numpy as np
import pymc3 as pm
import statsmodels.api as sm
import matplotlib.pyplot as plt
%matplotlib inline

data = pd.read_csv('ANES_2016_pilot.csv')
data.head()
data['vote'].value_counts()
data['party_id'].value_counts()

fig, ax = plt.subplots(3, figsize=(10,6))
key = dict(zip(data['party_id'].unique(),range(3)))
for label, df in data.groupby('party_id'):
    ax[key[label]].hist(df['age'])
    ax[key[label]].set_xlim([18,90])
    ax[key[label]].set_xlabel('Age')
    ax[key[label]].set_ylabel('Frequency')
    ax[key[label]].set_title(label)
    ax[key[label]].axvline(df['age'].mean())
plt.tight_layout()

pd.crosstab(data['vote'], data['party_id'])

clinton_data = data.loc[data['vote'].isin(['clinton','trump']),:]
clinton_data.head()

import bambi as bmb
clinton_model = bmb.Model(clinton_data)

clinton_model = bmb.Model(clinton_data)
clinton_fitted = clinton_model.fit('vote[clinton] ~ party_id + party_id:age', family="bernoulli", samples=1000, chains=4, init=None)

The error occurs after the above line. The full trace back is shown below.

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/pandas/core/frame.py:3140: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  self[k1] = value[k2]
/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/pandas/core/generic.py:4388: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.
  object.__getattribute__(self, name)
/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/pandas/core/generic.py:4389: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.
  return object.__setattr__(self, name, value)


---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-8-005c4346e6ee> in <module>()
      1 clinton_model = bmb.Model(clinton_data)
      2 clinton_fitted = clinton_model.fit('vote[clinton] ~ party_id + party_id:age',
----> 3     family='bernoulli', samples=1000, chains=4, init=None)

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/bambi/models.py in fit(self, fixed, random, priors, family, link, run, categorical, backend, **kwargs)
    278         if run:
    279             if not self.built or backend != self._backend_name:
--> 280                 self.build(backend)
    281             return self.backend.run(**kwargs)
    282 

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/bambi/models.py in build(self, backend)
    218                 taylor = 5 if self.family.name == 'gaussian' else 1
    219             scaler = PriorScaler(self, taylor=taylor)
--> 220             scaler.scale()
    221 
    222         # For bernoulli models with n_trials = 1 (most common use case),

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/bambi/priors.py in scale(self)
    409 
    410             # scale it!
--> 411             getattr(self, '_scale_%s' % term_type)(t)

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/bambi/priors.py in _scale_fixed(self, term)
    308             mu += [0]
    309             sd += [self._get_slope_stats(exog=self.dm, predictor=pred,
--> 310                                          sd_corr=sd_corr)]
    311 
    312         # save and set prior

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/bambi/priors.py in _get_slope_stats(self, exog, predictor, sd_corr, full_mod, points)
    228                                 str(exog.columns[i])+'='+str(val),
    229                                 start_params=full_mod.params.values)
--> 230                     for val in values[:-1]]
    231             null = np.append(null, full_mod)
    232             ll = np.array([x.llf for x in null])

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/bambi/priors.py in <listcomp>(.0)
    228                                 str(exog.columns[i])+'='+str(val),
    229                                 start_params=full_mod.params.values)
--> 230                     for val in values[:-1]]
    231             null = np.append(null, full_mod)
    232             ll = np.array([x.llf for x in null])

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/statsmodels/genmod/generalized_linear_model.py in fit_constrained(self, constraints, start_params, **fit_kwds)
   1284         params, cov, res_constr = fit_constrained(self, R, q,
   1285                                                   start_params=start_params,
-> 1286                                                   fit_kwds=fit_kwds)
   1287         # create dummy results Instance, TODO: wire up properly
   1288         res = self.fit(start_params=params, maxiter=0)  # we get a wrapper back

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/statsmodels/base/_constraints.py in fit_constrained(model, constraint_matrix, constraint_values, start_params, fit_kwds)
    258     # using offset as keywords is not supported in all modules
    259     mod_constr = self.__class__(endog, exogp_st, offset=offset, **init_kwds)
--> 260     res_constr = mod_constr.fit(start_params=start_params, **fit_kwds)
    261     params_orig = transf.expand(res_constr.params).squeeze()
    262     cov_params = transf.transf_mat.dot(res_constr.cov_params()).dot(transf.transf_mat.T)

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/statsmodels/genmod/generalized_linear_model.py in fit(self, start_params, maxiter, method, tol, scale, cov_type, cov_kwds, use_t, full_output, disp, max_start_irls, **kwargs)
   1010             return self._fit_irls(start_params=start_params, maxiter=maxiter,
   1011                                   tol=tol, scale=scale, cov_type=cov_type,
-> 1012                                   cov_kwds=cov_kwds, use_t=use_t, **kwargs)
   1013         else:
   1014             self._optim_hessian = kwargs.get('optim_hessian')

/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/statsmodels/genmod/generalized_linear_model.py in _fit_irls(self, start_params, maxiter, tol, scale, cov_type, cov_kwds, use_t, **kwargs)
   1107                                    self.freq_weights, self.scale)
   1108         if np.isnan(dev):
-> 1109             raise ValueError("The first guess on the deviance function "
   1110                              "returned a nan.  This could be a boundary "
   1111                              " problem and should be reported.")

ValueError: The first guess on the deviance function returned a nan.  This could be a boundary  problem and should be reported.

[MooersLab]

Hi Osvaldo,

Thank you very much your updating bambi and responding to this four month old open issue. I too found the notebook in question to be working today with Python 3.5 and Python 3.7 from macports.

I use anaconda (it is getting better all of the time), but I still prefer macports (faster installs, more packages, fewer dependency conflicts). The following may only interest other macports users. I had updated pymc3 through macports. This action updated the joblib module that pymc3 depends on to a version that is newer than what Bambi accepts. I had to uninstall joblib and pymc3 from my macports distribution. Then I installed Bambi with pip from the GitHub repository. Then I installed pymc3 with pip from the GitHub repository. I then found that the notebook worked with both python3.5 and python3.7.

On Sun, Mar 31, 2019 at 8:11 AM Osvaldo Martin [e-mail] wrote:

I am not able to reproduce this. Is this still a problem?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bambinos/bambi/issues/129#issuecomment-478340588, or mute the thread https://github.com/notifications/unsubscribe-auth/AOeSC6Qkedguig5bFJ36xjCR4oN4zzLHks5vcLQQgaJpZM4ZCCN0 .

– Best regards,

Blaine

[MooersLab]

Time to close this issue.