def test_Spatial_Median_Rate(self):
out_smr = sm.Spatial_Median_Rate(self.e, self.b, self.w).r
out_smr_w = sm.Spatial_Median_Rate(self.e, self.b, self.w, aw=self.b).r
out_smr2 = sm.Spatial_Median_Rate(
self.e, self.b, self.w, iteration=2).r
out_smr = [round(i, 7) for i in out_smr[:5]]
out_smr_w = [round(i, 7) for i in out_smr_w[:5]]
out_smr2 = [round(i, 7) for i in out_smr2[:5]]
self.assertEquals(out_smr, self.smr)
self.assertEquals(out_smr_w, self.smr_w)
self.assertEquals(out_smr2, self.smr2)
def test_Spatial_Median_Rate(self):
out_smr = sm.Spatial_Median_Rate(self.e, self.b, self.w).r
out_smr_w = sm.Spatial_Median_Rate(self.e, self.b, self.w, aw=self.b).r
out_smr2 = sm.Spatial_Median_Rate(self.e, self.b, self.w,
iteration=2).r
np.testing.assert_allclose(out_smr[:5].flatten(),
self.smr,
atol=ATOL,
rtol=RTOL)
np.testing.assert_allclose(out_smr_w[:5].flatten(),
self.smr_w,
atol=ATOL,
rtol=RTOL)
np.testing.assert_allclose(out_smr2[:5].flatten(),
self.smr2,
atol=ATOL,
rtol=RTOL)
def test_Spatial_Median_Rate_tabular(self):
out_smr = sm.Spatial_Median_Rate(self.df[self.ename],
self.df[self.bname], self.w).r
out_smr_w = sm.Spatial_Median_Rate(self.df[self.ename],
self.df[self.bname],
self.w,
aw=self.df[self.bname]).r
out_smr2 = sm.Spatial_Median_Rate(self.df[self.ename],
self.df[self.bname],
self.w,
iteration=2).r
self.assertIsInstance(out_smr, np.ndarray)
self.assertIsInstance(out_smr_w, np.ndarray)
self.assertIsInstance(out_smr2, np.ndarray)
np.testing.assert_allclose(out_smr[:5].flatten(),
self.smr,
atol=ATOL,
rtol=RTOL)
np.testing.assert_allclose(out_smr_w[:5].flatten(),
self.smr_w,
atol=ATOL,
rtol=RTOL)
np.testing.assert_allclose(out_smr2[:5].flatten(),
self.smr2,
atol=ATOL,
rtol=RTOL)
out_smr = sm.Spatial_Median_Rate.by_col(self.df, self.ename,
self.bname, self.w)
out_smr_w = sm.Spatial_Median_Rate.by_col(self.df,
self.ename,
self.bname,
self.w,
aw=self.df[self.bname])
out_smr2 = sm.Spatial_Median_Rate.by_col(self.df,
self.ename,
self.bname,
self.w,
iteration=2)
outcol = '{}-{}_spatial_median_rate'.format(self.ename, self.bname)
np.testing.assert_allclose(out_smr[outcol].values[:5],
self.smr,
rtol=RTOL,
atol=ATOL)
np.testing.assert_allclose(out_smr_w[outcol].values[:5],
self.smr_w,
rtol=RTOL,
atol=ATOL)
np.testing.assert_allclose(out_smr2[outcol].values[:5],
self.smr2,
rtol=RTOL,
atol=ATOL)
@unittest.skipIf(PANDAS_EXTINCT, 'missing pandas')
def test_Spatial_Smoother_multicol(self):
"""
test that specifying multiple columns works correctly. Since the
function is shared over all spatial smoothers, we can only test one.
"""
enames = [self.ename, 'SID79']
bnames = [self.bname, 'BIR79']
out_df = sm.Spatial_Median_Rate.by_col(self.df, enames, bnames,
self.w)
outcols = [
'{}-{}_spatial_median_rate'.format(e, b)
for e, b in zip(enames, bnames)
]
smr79 = np.array(
[0.00122129, 0.00176924, 0.00176924, 0.00240964, 0.00272035])
answers = [self.smr, smr79]
for col, answer in zip(outcols, answer):
self.assertIn(out_df.columns, col)
np.testing.assert_allclose(out_df[col].values[:5],
answer,
rtol=RTOL,
atol=ATOL)
@unittest.skipIf(PANDAS_EXTINCT, 'missing pandas')
def test_Smoother_multicol(self):
"""
test that non-spatial smoothers work with multicolumn queries
"""
enames = [self.ename, 'SID79']
bnames = [self.bname, 'BIR79']
out_df = sm.Excess_Risk.by_col(self.df, enames, bnames)
outcols = [
'{}-{}_excess_risk'.format(e, b)
for e, b in zip(enames, bnames)
]
er79 = np.array(
[0.000000, 2.796607, 0.8383863, 1.217479, 0.943811])
answers = [self.er, er79]
for col, answer in zip(outcols, answer):
self.assertIn(out_df.columns, col)
np.testing.assert_allclose(out_df[col].values[:5],
answer,
rtol=RTOL,
atol=ATOL)
elif sigs[i] > 0.0:
sigs[i] = 4
# plot significant autocorrelation
maps.plot_choropleth(shp_link, np.array(sigs), type='equal_interval',
title='Significant Spatial Autocorrelation in 311 Calls', k=10, figsize=(6,9))
# spatial smoothing
from pysal.esda import smoothing as sm
pop = np.array(calls['pop'])
all311 = np.array(calls['all_calls'])
# Locally weighted median smoothing, weighted by pop
rate = sm.Spatial_Median_Rate(all311, pop, w, aw=pop)
# weights are populations
y = rate.r
# check for nans
len(y[np.isnan(y)])
y[np.isnan(y)] = 0
maps.plot_choropleth(shp_link, y, type='fisher_jenks',
title='All Calls by Census Area, 2011-2015\nSpatially Smoothed',
k=20, figsize=(6,9))
# Locally weighted median smoothing, weighted by pop, iterated three times
rate = sm.Spatial_Median_Rate(all311, pop, w, aw=pop, iteration=3)
# weights are populations
y = rate.r