def fit(self):
df = self.data.copy()
df['group'] = 'All'
# smooth
self.smoothed = neighbor_mean_std(
df=df, col_val=self.outcome,
col_group='group',
col_axis=self.covariates,
radius=self.radius
)
if self.num_smooths > 1:
i = 1
data = self.smoothed.copy()
data.rename(columns={'residual_std': 'residual'}, inplace=True)
data.drop(columns={'residual_mean'}, axis=1, inplace=True)
while i < self.num_smooths:
self.smoothed = neighbor_mean_std(
df=data, col_val='residual',
col_group='group',
col_axis=self.covariates,
radius=self.radius
)
data = self.smoothed.copy()
data.rename(columns={'residual_mean': 'residual'}, inplace=True)
data.drop(['residual_std'], inplace=True, axis=1)
i += 1
self.smoothed.drop('residual_std', inplace=True, axis=1)
self.smoothed.rename(columns={'residual_mean': 'residual_std'}, inplace=True)
def test_neighbor_mean_std(testing_problem):
data = testing_problem
my_result = utils.neighbor_mean_std(data,
col_axis=['far_out', 'num_data'],
col_val='residual',
col_group='group',
radius=[2, 2])
cols = ['far_out', 'num_data', 'group', 'residual_mean', 'residual_std']
assert all([c in my_result for c in cols])
def fit(self):
df = self.data.copy()
df['Smooth_Group'] = 'All'
# smooth
self.smoothed = neighbor_mean_std(df=df,
col_val=self.outcome,
col_group='Smooth_Group',
col_axis=self.covariates,
radius=self.radius)
def get_smoothed_residuals(self, radius):
"""
Smooth residuals for all of the data across some radius
Args:
radius: List[int] with two elements indicating the width and height
of the square that we want to smooth over.
Returns:
smoothed_residuals: (pd.DataFrame)
"""
smoothed_residuals = neighbor_mean_std(
df=self.all_residuals,
col_val='residual',
col_group=self.col_group,
col_axis=['far_out', 'num_data'],
radius=radius
)
return smoothed_residuals