def test_Kernel(self):
kw = pysal.Kernel(self.points)
self.assertEqual(kw.weights[0],
[1.0, 0.50000004999999503, 0.44098306152674649])
kw15 = pysal.Kernel(self.points, bandwidth=15.0)
self.assertEqual(kw15[0], {
0: 1.0,
1: 0.33333333333333337,
3: 0.2546440075000701
})
self.assertEqual(kw15.bandwidth[0], 15.)
self.assertEqual(kw15.bandwidth[-1], 15.)
bw = [25.0, 15.0, 25.0, 16.0, 14.5, 25.0]
kwa = pysal.Kernel(self.points, bandwidth=bw)
self.assertEqual(kwa.weights[0], [
1.0, 0.59999999999999998, 0.55278640450004202, 0.10557280900008403
])
self.assertEqual(kwa.neighbors[0], [0, 1, 3, 4])
self.assertEqual(kwa.bandwidth[0], 25.)
self.assertEqual(kwa.bandwidth[1], 15.)
self.assertEqual(kwa.bandwidth[2], 25.)
self.assertEqual(kwa.bandwidth[3], 16.)
self.assertEqual(kwa.bandwidth[4], 14.5)
self.assertEqual(kwa.bandwidth[5], 25.)
kwea = pysal.Kernel(self.points, fixed=False)
self.assertEqual(kwea.weights[0],
[1.0, 0.10557289844279438, 9.9999990066379496e-08])
l = kwea.bandwidth.tolist()
self.assertEqual(
l,
[[11.180341005532938], [11.180341005532938], [20.000002000000002],
[11.180341005532938], [14.142137037944515], [18.027758180095585]])
kweag = pysal.Kernel(self.points, fixed=False, function='gaussian')
self.assertEqual(
kweag.weights[0],
[0.3989422804014327, 0.26741902915776961, 0.24197074871621341])
l = kweag.bandwidth.tolist()
self.assertEqual(
l,
[[11.180341005532938], [11.180341005532938], [20.000002000000002],
[11.180341005532938], [14.142137037944515], [18.027758180095585]])
kw = pysal.kernelW_from_shapefile(self.polyShp, idVariable='POLYID')
self.assertEqual(
set(kw.weights[1]),
set([
0.0070787731484506233, 0.2052478782400463, 0.23051223027663237,
1.0
]))
kwa = pysal.adaptive_kernelW_from_shapefile(self.polyShp)
self.assertEqual(kwa.weights[0],
[1.0, 0.03178906767736345, 9.9999990066379496e-08])
def main():
""" Reports the spatial autocorrelation for each target variable. """
ap = argparse.ArgumentParser()
ap.add_argument(
'--coords',
default='/data/twcounty/stats2/counties/counties.top100.bounding.txt',
help='tsv file with counties, lat/long, and bounding boxes')
ap.add_argument(
'--dbf',
default=
'/data/twcounty/stats2/counties/census_polygons/gz_2010_us_050_00_500k.dbf',
help='county dbf files')
ap.add_argument(
'--input',
default='/data/twcounty/features/counties.norm=none.liwc.pkl',
help='pickled county feature Data object')
ap.add_argument('--json',
default='/data/twcounty/json',
help='directory of jsons per county')
ap.add_argument(
'--shapes',
default=
'/data/twcounty/stats2/counties/census_polygons/gz_2010_us_050_00_500k.shp',
help='county shape files')
ap.add_argument('--targets',
default='/data/twcounty/targets.tsv',
help='targets per county')
ap.add_argument('--tweets',
default='/data/twcounty/tweets.tsv',
help='tweets')
args = ap.parse_args()
coords = read_coords(args.coords)
overlaps = get_overlaps(coords)
print_tweets_in_overlaps(overlaps, args.json, args.tweets)
shapes = pysal.open(args.shapes)
print 'read', len(shapes), 'counties from shape file'
dbf = pysal.open(args.dbf)
counties = data.read(args.input).features
targets, target_alpha = read_targets(args.targets, counties)
for t in targets:
didx, d = [(i, dd) for i, dd in enumerate(dbf)
if dd[1] + dd[2] == t][0]
targets[t]['dbf'] = d
targets[t]['dbfi'] = didx
shape = shapes[didx]
targets[t]['shape'] = shape
fips = sorted(targets)
print '\t'.join(['mi.I ', 'mi.EI ', 'p_norm ', 'outcome'])
weights = pysal.Kernel(
np.array([targets[f]['shape'].centroid for f in fips]))
for outcome in target_alpha:
y = np.array([float(targets[t][outcome]) for t in fips])
y = y / np.sum(y)
mi = pysal.Moran(y, weights)
print '%.5f\t%.5f\t%.5f\t%s' % (mi.I, mi.EI, mi.p_norm, outcome)
def make_distance_weights():
""" Compute a distance weight matrix based on the geographical coordinates
of each county. """
coords = twexpt.read_coords(
'/data/twcounty/stats2/counties/counties.top100.bounding.txt')
coords = [coords[cty] for cty in sorted(coords)]
weights = pysal.Kernel(np.array(coords))
for i in range(len(weights.weights)):
idx = weights.neighbors[i].index(i)
del weights.weights[i][idx]
del weights.neighbors[i][idx]
weights.transform = 'r'
return weights
def gpd_kernel(gpdf,
bandwidth=None,
fixed=True,
k=2,
diagonal=False,
function='triangular',
eps=1.0000001,
ids=None):
"""
Kernel weights
https://github.com/pysal/pysal/blob/master/pysal/weights/Distance.py
"""
points = gpd_points(gpdf)
kw = pysal.Kernel(points,
bandwidth=bandwidth,
fixed=fixed,
k=k,
diagonal=diagonal,
function=function,
eps=eps,
ids=ids)
return kw