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
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
|
"""
Returns interpolated scores (values) at percentiles through a sample
"""
from __future__ import print_function
import numpy
import pyferret
import scipy.stats
def ferret_init(id):
"""
Initialization for the stats_scoresatpercentiles.py Ferret PyEF
"""
axes_values = [ pyferret.AXIS_IMPLIED_BY_ARGS ] * pyferret.MAX_FERRET_NDIM
true_influences = [ True ] * pyferret.MAX_FERRET_NDIM
false_influences = [ False ] * pyferret.MAX_FERRET_NDIM
retdict = { "numargs": 2,
"descript": "Returns interpolated scores (values) that are given percentiles through a sample",
"axes": axes_values,
"argnames": ( "SAMPLE", "PERCENTILES", ),
"argdescripts": ( "Sample of scores (values)",
"Percentiles (0-100) through sample to find scores (values) of", ),
"argtypes": ( pyferret.FLOAT_ARRAY, pyferret.FLOAT_ARRAY, ),
"influences": ( false_influences, true_influences, ),
}
return retdict
def ferret_compute(id, result, resbdf, inputs, inpbdfs):
"""
Assigns result with interpolated scores that are given percentiles
through a sample. The sample scores are given in inputs[0], and the
percentiles are given in inputs[1]. Undefined values in inputs[0]
are eliminated before using it in scipy.stats.scoreatpercentile.
Undefined values in inputs[1] return corresponding undefined values
in result.
"""
# make sure result has the expected dimensions
if result.shape != inputs[1].shape:
raise ValueError("Unexpected error; PERCENTILE dimen: %s; result dimen: %s" % \
(str(inputs[1].shape), str(result.shape)))
# get the clean sample data as a flattened array
badmask = ( numpy.fabs(inputs[0] - inpbdfs[0]) < 1.0E-5 )
badmask = numpy.logical_or(badmask, numpy.isnan(inputs[0]))
goodmask = numpy.logical_not(badmask)
values = inputs[0][goodmask]
# get the mask for the good percentiles
badmask = ( numpy.fabs(inputs[1] - inpbdfs[1]) < 1.0E-5 )
badmask = numpy.logical_or(badmask, numpy.isnan(inputs[1]))
goodmask = numpy.logical_not(badmask)
# scoreatpercentile doesn't take an array for the percentiles
# so do them one at a time
scores = [ ]
for prcnt in inputs[1][goodmask]:
scores.append(scipy.stats.scoreatpercentile(values, prcnt))
result[goodmask] = scores
result[badmask] = resbdf
#
# The rest of this is just for testing this module at the command line
#
if __name__ == "__main__":
# make sure ferret_init does not have problems
info = ferret_init(0)
ydim = 10
zdim = 12
offset = 32.5
inpbdfs = numpy.array([-1.0, -2.0], dtype=numpy.float64)
resbdf = numpy.array([-3.0], dtype=numpy.float64)
sample = numpy.empty((1, ydim, zdim, 1, 1, 1), dtype=numpy.float64, order='F')
# valid sample values are [0:100:1] + offset
pval = 0
index = 0
for j in range(ydim):
for k in range(zdim):
if ((index % 7) == 3) or (pval > 100):
sample[0, j, k, 0, 0, 0] = inpbdfs[0]
else:
sample[0, j, k, 0, 0, 0] = pval + offset
pval += 1
index += 1
if pval != 101:
raise ValueError("Unexpected final pval of %d (ydim,zdim too small)" % pval)
prcnts = numpy.empty((1, 1, zdim, 1, 1, 1), dtype=numpy.float64, order='F')
expected = numpy.empty((1, 1, zdim, 1, 1, 1), dtype=numpy.float64, order='F')
prcnts[:,:,:,:,:,:] = inpbdfs[1]
expected[:,:,:,:,:,:] = resbdf
for k in ( 1, 2, 3, 5, 6, 7, 9 ):
prcnts[0, 0, k, 0, 0, 0] = 10.0 * k
expected[0, 0, k, 0, 0, 0] = 10.0 * k + offset
result = -888.0 * numpy.ones((1, 1, zdim, 1, 1, 1), dtype=numpy.float64, order='F')
ferret_compute(0, result, resbdf, (sample, prcnts), inpbdfs)
if not numpy.allclose(result, expected):
print("Expected (flattened) = %s" % str(expected.reshape(-1)))
print("Result (flattened) = %s" % str(result.reshape(-1)))
raise ValueError("Unexpected result")
# All successful
print("Success")
|
