ARMA start estimate with numpy master #957

Closed
josef-pkt opened this Issue Jul 8, 2013 · 3 comments

Projects

None yet

2 participants

@josef-pkt
Member

reported by Tom, when running the test suite with numpy and scipy masters

https://groups.google.com/d/msg/pystatsmodels/GBQYPHNwfSs/K9jaORVcX74J

several errors like the following

======================================================================
ERROR: test suite for <class 'statsmodels.tsa.tests.test_arima.Test_ARIMA111CSS'>
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/suite.py", line 208, in run
    self.setUp()
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/suite.py", line 291, in setUp
    self.setupContext(ancestor)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/suite.py", line 314, in setupContext
    try_run(context, names)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/util.py", line 469, in try_run
    return func()
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/tests/test_arima.py", line 587, in setupClass
    cls.res1 = ARIMA(cpi, (1,1,1)).fit(disp=-1, method='css')
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 997, in fit
    full_output, disp, callback, **kwargs)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 821, in fit
    start_params = self._fit_start_params((k_ar,k_ma,k), method)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 442, in _fit_start_params
    start_params = self._fit_start_params_hr(order)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 425, in _fit_start_params_hr
    raise ValueError("The computed initial AR coefficients are not "
ValueError: The computed initial AR coefficients are not stationary
You should induce stationarity, choose a different model order, or you can
pass your own start_params.

======================================================================
ERROR: test suite for <class 'statsmodels.tsa.tests.test_arima.Test_Y_ARMA11_Const'>
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/suite.py", line 208, in run
    self.setUp()
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/suite.py", line 291, in setUp
    self.setupContext(ancestor)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/suite.py", line 314, in setupContext
    try_run(context, names)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/nose/util.py", line 469, in try_run
    return func()
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/tests/test_arima.py", line 263, in setupClass
    cls.res1 = ARMA(endog, order=(1,1)).fit(trend="c", disp=-1)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 821, in fit
    start_params = self._fit_start_params((k_ar,k_ma,k), method)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 446, in _fit_start_params
    start_params = self._fit_start_params_hr(order)
  File "/Users/tom/python2.7/lib/python2.7/site-packages/statsmodels-0.5.0-py2.7-macosx-10.8-x86_64.egg/statsmodels/tsa/arima_model.py", line 425, in _fit_start_params_hr
    raise ValueError("The computed initial AR coefficients are not "
ValueError: The computed initial AR coefficients are not stationary
You should induce stationarity, choose a different model order, or you can
pass your own start_params.
@jseabold
Member

I can't replicate this with a newer version of numpy. I'm on 596795b, the version with this failure was from 3 weeks earlier b5fd429. Has anyone else reported this or have we seen it on any other system?

@jseabold
Member

It looks like after reinstalling numpy/scipy that these also went away. I'm inclined to close this if we haven't seen it elsewhere.

@josef-pkt
Member

closing, 0.5.0 is released

we can look at this again if it reappears

@josef-pkt josef-pkt closed this Aug 15, 2013
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment