def test_setup_labels_pi_none_nlines_2(self):
PI = PredictionIntervals()
clabels, plabels = PI._setup_labels(prediction_intervals=None,
nlines=2)
self.assertEqual(clabels, ['99% CI', '95% CI', '90% CI', '50% CI'],
msg='String should match')
self.assertEqual(plabels, ['95% PI'], msg='String should match')
def test_setup_counting_metrics(self):
PI = PredictionIntervals()
ci = gf.setup_pseudo_ci()
nbatch, nn, nlines = PI._setup_counting_metrics(credible_intervals=ci)
self.assertEqual(nbatch, 1, msg='Expect nbatch = 1')
self.assertEqual(nn, 2, msg='Expect nn = 2')
self.assertEqual(nlines, 1, msg='Expect nlines = 1')
def test_basic_ds_shape_0(self):
PI = PredictionIntervals()
DS = gf.basic_data_structure()
DS.shape = [(100, )] # remove column - code should add it back
ndatabatches, ncols = PI._analyze_data_structure(data=DS)
self.assertEqual(ndatabatches, 1, msg='Expect 1 batch')
self.assertEqual(ncols, [1], msg='Expect [1]')
def test_setup_int_lims_s2chain_none(self):
PI = PredictionIntervals()
lims = PI._setup_interval_limits(s2chain=None)
self.assertTrue(np.array_equal(
lims,
np.array([0.005, 0.025, 0.05, 0.25, 0.5, 0.75, 0.9, 0.975,
0.995])),
msg=str('Arrays should match: {}'.format(lims)))
def pi_setup():
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
DS = gf.basic_data_structure()
PI.setup_prediction_interval_calculation(results=results,
data=DS,
modelfunction=gf.predmodelfun)
return PI, results
def test_setup_sstype(self):
PI = PredictionIntervals()
PI._PredictionIntervals__sstype = 3
sstype = PI._setup_sstype(sstype=None)
self.assertEqual(sstype,
PI._PredictionIntervals__sstype,
msg='Expected 3')
sstype = PI._setup_sstype(sstype=2)
self.assertEqual(sstype, 0, msg='Expected 0')
def test_s2chain_index_n_neq_1_tc_neq_n_neq_ndatabatches(self):
PI = PredictionIntervals()
ndatabatches = 5
s2chain = np.random.random_sample(size=(100, 3))
ncol = [1, 2, 1]
with self.assertRaises(SystemExit, msg='Unrecognized data structure'):
PI._analyze_s2chain(ndatabatches=ndatabatches,
s2chain=s2chain,
ncol=ncol)
def test_define_sample_points_nsample_lte_nsimu(self, mock_rand):
PI = PredictionIntervals()
aa = np.random.rand([400, 1])
mock_rand.return_value = aa
iisample, nsample = PI._define_sample_points(nsample=400, nsimu=500)
self.assertTrue(np.array_equal(iisample,
np.ceil(aa * 500) - 1),
msg='Expect range(500)')
self.assertEqual(nsample, 400, msg='Expect nsample to stay 400')
def test_basic_datapred(self):
PI = PredictionIntervals()
DS = gf.basic_data_structure()
datapred = PI._setup_data_structure_for_prediction(data=DS,
ndatabatches=1)
self.assertTrue(np.array_equal(datapred[0].xdata[0], DS.xdata[0]),
msg='Arrays should match')
self.assertTrue(np.array_equal(datapred[0].ydata[0], DS.ydata[0]),
msg='Arrays should match')
def test_feature_assignment(self):
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
PI._assign_features_from_results_structure(results=results)
PID = PI.__dict__
check_these = [
'chain', 's2chain', 'parind', 'local', 'theta', 'sstype'
]
self.check_dictionary(check_these, PID, results)
def test_does_observation_sample_off_s2elem_greater_than_1_cause_system_exit(
self):
PI = PredictionIntervals()
s2elem = np.array([[2.0, 1.0]])
ypred = np.linspace(2.0, 3.0, num=15)
ypred = ypred.reshape(5, 3)
sstype = 0
with self.assertRaises(SystemExit):
PI._observation_sample(s2elem, ypred, sstype)
def test_second_feature_assignment(self):
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
results = gf.removekey(results, 'sstype')
PI._assign_features_from_results_structure(results=results)
PID = PI.__dict__
check_these = ['chain', 's2chain', 'parind', 'local', 'theta']
self.check_dictionary(check_these, PID, results)
self.assertEqual(PID['_PredictionIntervals__sstype'],
0,
msg='Should default to 0')
def test_s2chain_index_n_eq_1(self):
PI = PredictionIntervals()
ndatabatches = 1
s2chain = np.random.random_sample(size=(100, 1))
ncol = [1]
s2chain_index = PI._analyze_s2chain(ndatabatches=ndatabatches,
s2chain=s2chain,
ncol=ncol)
self.assertTrue(np.array_equal(s2chain_index, np.array([[0, 1]])),
msg=str(
'Arrays should match: {}'.format(s2chain_index)))
def test_does_observation_sample_wrong_size_s2elem_break_right_size_array(
self):
PI = PredictionIntervals()
s2elem = np.array([[2.0, 1.0]])
ypred = np.linspace(2.0, 3.0, num=5)
ypred = ypred.reshape(5, 1)
sstype = 0
with self.assertRaises(
SystemExit,
msg='Mismatched size for s2chain and model return'):
PI._observation_sample(s2elem, ypred, sstype)
def test_non_basic_modelfunction(self):
PI = PredictionIntervals()
DS = gf.non_basic_data_structure()
datapred = PI._setup_data_structure_for_prediction(data=DS,
ndatabatches=2)
nrow, ncol = PI._determine_shape_of_response(
modelfunction=gf.predmodelfun,
ndatabatches=2,
datapred=datapred,
theta=[3.0, 5.0])
self.assertEqual(nrow, [100, 100], msg='Expect [100, 100]')
self.assertEqual(ncol, [1, 1], msg='Expect [1, 1]')
def test_does_observation_sample_unknown_sstype_cause_system_exit(self):
PI = PredictionIntervals()
s2elem = np.array([[2.0]])
ypred = np.linspace(2.0, 3.0, num=5)
ypred = ypred.reshape(5, 1)
sstype = 4
with self.assertRaises(SystemExit, msg='Unrecognized sstype'):
PI._observation_sample(s2elem, ypred, sstype)
sstype = 0
opred = PI._observation_sample(s2elem, ypred, sstype)
self.assertEqual(opred.shape, ypred.shape, msg='Shapes should match')
def test_add_batch_column_title_ny_gt_1(self):
PI = PredictionIntervals()
htmp, ax = PI._initialize_plot_features(ii=0,
jj=0,
ny=1,
figsizeinches=[10, 11])
PI._add_batch_column_title(nbatch=1, ny=2, ii=0, jj=0)
self.assertEqual(ax.get_title(),
'Column #0',
msg='Strings should match')
self.assertEqual(htmp.get_figwidth(), 10.0, msg='Expect 10.0')
self.assertEqual(htmp.get_figheight(), 11.0, msg='Expect 11.0')
def test_initialize_plot_features(self):
PI = PredictionIntervals()
htmp, ax = PI._initialize_plot_features(ii=0,
jj=0,
ny=1,
figsizeinches=[10, 11])
self.assertEqual(htmp.get_figwidth(), 10.0, msg='Figure width is 10in')
self.assertEqual(htmp.get_figheight(),
11.0,
msg='Figure height is 11in')
self.assertEqual(htmp.get_label(),
'Batch # 0 | Column # 0',
msg='Strings should match')
self.assertEqual(ax.get_label(), '', msg='Expect empty string')
def check_these_settings(self, infigsizeinches, expected):
PI = PredictionIntervals()
ci = gf.setup_pseudo_ci()
prediction_intervals, figsizeinches, nbatch, nn, clabels, plabels = PI._setup_interval_plotting(
plot_pred_int=True,
prediction_intervals=None,
credible_intervals=ci,
figsizeinches=infigsizeinches)
self.assertEqual(figsizeinches,
expected,
msg=str('Expect {}'.format(expected)))
self.assertEqual(prediction_intervals, None, msg='Expect None')
self.assertEqual(nbatch, 1, msg='Expect nbatch = 1')
self.assertEqual(nn, 2, msg='Expect nn = 2')
self.assertEqual(clabels, ['95% CI'], msg='String should match')
self.assertEqual(plabels, ['95% PI'], msg='String should match')
def test_setup_display_settings(self):
PI = PredictionIntervals()
model_display = {'label': 'hello'}
data_display = {'linewidth': 7}
interval_display = {'edgecolor': 'b'}
intd, modd, datd = PI._setup_display_settings(
interval_display=interval_display,
model_display=model_display,
data_display=data_display)
self.assertEqual(modd['label'],
model_display['label'],
msg='Expect label to match')
self.assertEqual(intd['edgecolor'],
interval_display['edgecolor'],
msg='Expect edgecolor to match')
self.assertEqual(datd['linewidth'],
data_display['linewidth'],
msg='Expect linewidth to match')
def test_setup_pi_calc_with_s2chain_none(self):
PI = PredictionIntervals()
DS = gf.basic_data_structure()
results = gf.setup_pseudo_results()
results['s2chain'] = None
PI.setup_prediction_interval_calculation(results=results,
data=DS,
modelfunction=gf.predmodelfun)
self.assertEqual(PI._PredictionIntervals__ndatabatches,
1,
msg='Expect ndatabatches = 1')
self.assertEqual(PI.modelfunction,
gf.predmodelfun,
msg='Functions should match')
self.assertFalse(hasattr(PI, '_PredictionIntervals__s2chain_index'),
msg='Expect False')
self.assertTrue(np.array_equal(PI._PredictionIntervals__chain,
results['chain']),
msg='Arrays should match')
def cc_setup(calc_pred_int=True):
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
PI._assign_features_from_results_structure(results=results)
testchain, s2chain, lims, sstype, nsample, iisample = PI._setup_generation_requirements(
nsample=400, calc_pred_int=calc_pred_int, sstype=0)
DS = gf.basic_data_structure()
datapred = PI._setup_data_structure_for_prediction(data=DS, ndatabatches=1)
return PI, results, testchain, s2chain, lims, sstype, nsample, iisample, datapred
def setup_complex_pi(cls):
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
results['s2chain'] = None
DS = gf.basic_data_structure()
PI.setup_prediction_interval_calculation(results=results,
data=DS,
modelfunction=gf.predmodelfun)
PI.generate_prediction_intervals(sstype=0,
nsample=500,
calc_pred_int=False,
waitbar=False)
return PI
def test_generate_credible_intervals(self):
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
results['s2chain'] = None
DS = gf.basic_data_structure()
PI.setup_prediction_interval_calculation(results=results,
data=DS,
modelfunction=gf.predmodelfun)
PI.generate_prediction_intervals(sstype=None,
nsample=500,
calc_pred_int=False,
waitbar=False)
cint = PI.intervals['credible_intervals']
pint = PI.intervals['prediction_intervals']
self.common_set_1(cint, pint)
def test_generate_credible_intervals_with_waitbar(self):
PI = PredictionIntervals()
results = gf.setup_pseudo_results()
results['s2chain'] = None
DS = gf.basic_data_structure()
PI.setup_prediction_interval_calculation(results=results,
data=DS,
modelfunction=gf.predmodelfun)
PI.generate_prediction_intervals(sstype=None,
nsample=500,
calc_pred_int=False,
waitbar=True)
cint = PI.intervals['credible_intervals']
pint = PI.intervals['prediction_intervals']
self.common_set_1(cint, pint)
self.assertEqual(PI._PredictionIntervals__wbarstatus.percentage(3),
3.0,
msg='Expect 3.0')
def test_does_observation_sample_2_column_ypred_right_size_array(self):
PI = PredictionIntervals()
ypred = np.linspace(2.0, 3.0, num=10)
ypred = ypred.reshape(5, 2)
s2elem = np.array([[2.0, 1.0]])
sstype = 0
opred = PI._observation_sample(s2elem, ypred, sstype)
self.assertEqual(opred.shape, ypred.shape, msg='Shapes are compatible')
s2elem = np.array([[2.0]])
opred = PI._observation_sample(s2elem, ypred, sstype)
self.assertEqual(opred.shape, ypred.shape, msg='Shapes are compatible')
sstype = 1
opred = PI._observation_sample(s2elem, ypred, sstype)
self.assertEqual(opred.shape, ypred.shape, msg='Shapes are compatible')
sstype = 2
opred = PI._observation_sample(s2elem, ypred, sstype)
self.assertEqual(opred.shape, ypred.shape, msg='Shapes are compatible')
def test_ds_2_nbatch(self):
PI = PredictionIntervals()
DS = gf.non_basic_data_structure()
ndatabatches, ncols = PI._analyze_data_structure(data=DS)
self.assertEqual(ndatabatches, 2, msg='Expect 2 batches')
self.assertEqual(ncols, [1, 2], msg='Expect [1, 1]')
def test_basic_ds(self):
PI = PredictionIntervals()
DS = gf.basic_data_structure()
ndatabatches, ncols = PI._analyze_data_structure(data=DS)
self.assertEqual(ndatabatches, 1, msg='Expect 1 batch')
self.assertEqual(ncols, [1], msg='Expect [1]')
def setup_interval(cls):
PI = PredictionIntervals()
DS = gf.basic_data_structure()
datapred = PI._setup_data_structure_for_prediction(data=DS,
ndatabatches=1)
return PI, datapred
def gq_settings():
PI = PredictionIntervals()
nsample = 500
nrow = 100
ncol = 1
return PI, nsample, nrow, ncol
