1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
|
Description: Avoid NaN crash in test_structural
Out-of-range frequency.cycle starting values make untransform_params return NaN,
which crashes summary()
Author: Rebecca N. Palmer <[email protected]>
Forwarded: no
--- a/statsmodels/tsa/statespace/tests/results/results_structural.py
b/statsmodels/tsa/statespace/tests/results/results_structural.py
@@ -291,12 291,15 @@ lltrend_cycle_seasonal_reg_ar1_approx_di
# Monthly frequency dataset
{'level': 'lltrend', 'autoregressive': 1, 'cycle': True,
'stochastic_cycle': True, 'seasonal': 4, 'autoregressive': 1,
- 'exog': True, 'mle_regression': False, 'freq': 'MS'},
'exog': True, 'mle_regression': False, 'freq': 'MS',
'cycle_period_bounds': (1.5*4, 12*4)},
# explicit cycle_period_bounds needed because the default *12
# makes start_params out of range
# Minutely frequency dataset
{'level': 'lltrend', 'autoregressive': 1, 'cycle': True,
'stochastic_cycle': True, 'seasonal': 4, 'autoregressive': 1,
'exog': True, 'mle_regression': False, 'freq': 'T',
- 'cycle_period_bounds': (1.5*12, 12*12)},
'cycle_period_bounds': (1.5*4, 12*4)},
],
'params': [0.0001, 0.01, 0.06, 0.0001, 0.0001, 0.1, 2*pi / 10, 0.2],
'start_params': [0.0001, 0.01, 0.06, 0.0001, 0.0001, 0.1, 2*pi / 10, 0.2],
|