Develop

torchcast/__init__.py

@@ -1 +1 @@
-__version__ = '0.5.0'
+__version__ = '0.5.1'

torchcast/kalman_filter/binomial_filter.py

@@ -10,11 +10,11 @@
 
 from torchcast.covariance import Covariance
 from torchcast.covariance.util import mini_cov_mask
-from torchcast.internals.utils import identity
+from torchcast.internals.utils import identity, class_or_instancemethod
 from torchcast.kalman_filter import KalmanFilter
 from torchcast.kalman_filter.ekf import EKFStep, EKFPredictions
 from torchcast.process import Process
-from torchcast.utils import conf2bounds, class_or_instancemethod
+from torchcast.state_space.predictions import conf2bounds
 
 if TYPE_CHECKING:
     from pandas import DataFrame

torchcast/kalman_filter/ekf.py

@@ -6,8 +6,9 @@
 from torch import Tensor
 
 from .kalman_filter import KalmanStep
+from ..internals.utils import class_or_instancemethod
 from ..state_space import Predictions
-from ..utils import TimeSeriesDataset, class_or_instancemethod
+from ..utils import TimeSeriesDataset
 
 
 class EKFStep(KalmanStep):

torchcast/state_space/predictions.py

@@ -463,7 +463,11 @@ def _tensor_to_df(tens, measures):
             raise ValueError("Expected `type` to be 'predictions' or 'components'.")
 
         out = pd.concat(out).reset_index(drop=True)
-        _out_cols = [group_colname, time_colname, 'measure', 'mean', 'lower', 'upper']
+        _out_cols = [group_colname, time_colname, 'measure', 'mean']
+        if conf is None:
+            _out_cols.append('std')
+        else:
+            _out_cols.extend(['lower', 'upper'])
         if type.startswith('comp'):
             _out_cols = _out_cols[0:3] + ['process', 'state_element'] + _out_cols[3:]
         if 'actual' in out.columns: