def test_confusion_matrix_from_data(self):
"""confusion_matrix_from_data should work for binary and quantitative data"""
#binary test
exp = [1,1,1,0,0,0,1,1,1,0,1,0,1]
obs = [1,1,0,1,0,0,1,1,1,1,1,1,0]
exp_tp = 6.0
exp_fp = 3.0
exp_fn = 2.0
exp_tn = 2.0
tp,fp,fn,tn = confusion_matrix_from_data(obs,exp)
self.assertEqual([tp,fp,fn,tn],\
[exp_tp,exp_fp,exp_fn,exp_tn])
#quantitative test
#results should be indentical to binary test
#since only presence/absence is considered
exp = [1,1,1,0,0,0,1,1,1,0,1,0,1]
obs = [13.7,6.5,0,1,0,0,2.3,1,1.0,1.3,1.5,1,0]
exp_tp = 6.0
exp_fp = 3.0
exp_fn = 2.0
exp_tn = 2.0
tp,fp,fn,tn = confusion_matrix_from_data(obs,exp)
self.assertEqual([tp,fp,fn,tn],\
[exp_tp,exp_fp,exp_fn,exp_tn])
def test_confusion_matrix_from_data(self):
"""confusion_matrix_from_data should work for binary and quantitative data"""
#binary test
exp = [1,1,1,0,0,0,1,1,1,0,1,0,1]
obs = [1,1,0,1,0,0,1,1,1,1,1,1,0]
exp_tp = 6.0
exp_fp = 3.0
exp_fn = 2.0
exp_tn = 2.0
tp,fp,fn,tn = confusion_matrix_from_data(obs,exp)
self.assertEqual([tp,fp,fn,tn],\
[exp_tp,exp_fp,exp_fn,exp_tn])
#quantitative test
#results should be indentical to binary test
#since only presence/absence is considered
exp = [1,1,1,0,0,0,1,1,1,0,1,0,1]
obs = [13.7,6.5,0,1,0,0,2.3,1,1.0,1.3,1.5,1,0]
exp_tp = 6.0
exp_fp = 3.0
exp_fn = 2.0
exp_tn = 2.0
tp,fp,fn,tn = confusion_matrix_from_data(obs,exp)
self.assertEqual([tp,fp,fn,tn],\
[exp_tp,exp_fp,exp_fn,exp_tn])
def test_roc_points(self):
"""roc_points should calculate the points for a Receiver Operating Characteristics curve
"""
#The set up here is a bit elaborate since I generate the test datasets
#based on the values we need in the confusion matrix.
#I test the intermediate results though, so any errors should be due
#to the actual function, not the test
tn_obs = 0
tn_exp = 0
fp_obs = 1
fp_exp = 0
tp_obs = 1
tp_exp = 1
fn_obs = 0
fn_exp = 1
#point A
obs = [tp_obs] * 63 + [fp_obs] *28 + [fn_obs] * 37 + [tn_obs]*72
exp = [tp_exp] * 63 + [fp_exp] *28 + [fn_exp] * 37 + [tn_exp]*72
trial_a_results = confusion_matrix_from_data(obs,exp)
#Check that this is correct
self.assertEqual(trial_a_results,(63,28,37,72))
trial_a = (obs,exp)
#point B
obs = [tp_obs] * 77 + [fp_obs] *77 + [fn_obs] * 23 + [tn_obs]*23
exp = [tp_exp] * 77 + [fp_exp] *77 + [fn_exp] * 23 + [tn_exp]*23
trial_b_results = confusion_matrix_from_data(obs,exp)
#Check that this is correct
self.assertEqual(trial_b_results,(77,77,23,23))
trial_b = (obs,exp)
#point c
obs = [tp_obs] * 24 + [fp_obs] *88 + [fn_obs] * 76 + [tn_obs]*12
exp = [tp_exp] * 24 + [fp_exp] *88 + [fn_exp] * 76 + [tn_exp]*12
trial_c_results = confusion_matrix_from_data(obs,exp)
#Check that this is correct
self.assertEqual(trial_c_results,(24,88,76,12))
trial_c_results = calculate_accuracy_stats_from_observations(obs,exp)
#Check that this is correct
self.assertFloatEqual(trial_c_results["false_positive_rate"],0.88)
trial_c = (obs,exp)
trials = [trial_a, trial_b,trial_c]
#Finally the actual test
obs_points = roc_points(trials)
exp_points = [(0.28,0.63),(0.77,0.77),(0.88,0.24)]
self.assertFloatEqual(obs_points,exp_points)
def test_roc_points(self):
"""roc_points should calculate the points for a Receiver Operating Characteristics curve
"""
#The set up here is a bit elaborate since I generate the test datasets
#based on the values we need in the confusion matrix.
#I test the intermediate results though, so any errors should be due
#to the actual function, not the test
tn_obs = 0
tn_exp = 0
fp_obs = 1
fp_exp = 0
tp_obs = 1
tp_exp = 1
fn_obs = 0
fn_exp = 1
#point A
obs = [tp_obs] * 63 + [fp_obs] *28 + [fn_obs] * 37 + [tn_obs]*72
exp = [tp_exp] * 63 + [fp_exp] *28 + [fn_exp] * 37 + [tn_exp]*72
trial_a_results = confusion_matrix_from_data(obs,exp)
#Check that this is correct
self.assertEqual(trial_a_results,(63,28,37,72))
trial_a = (obs,exp)
#point B
obs = [tp_obs] * 77 + [fp_obs] *77 + [fn_obs] * 23 + [tn_obs]*23
exp = [tp_exp] * 77 + [fp_exp] *77 + [fn_exp] * 23 + [tn_exp]*23
trial_b_results = confusion_matrix_from_data(obs,exp)
#Check that this is correct
self.assertEqual(trial_b_results,(77,77,23,23))
trial_b = (obs,exp)
#point c
obs = [tp_obs] * 24 + [fp_obs] *88 + [fn_obs] * 76 + [tn_obs]*12
exp = [tp_exp] * 24 + [fp_exp] *88 + [fn_exp] * 76 + [tn_exp]*12
trial_c_results = confusion_matrix_from_data(obs,exp)
#Check that this is correct
self.assertEqual(trial_c_results,(24,88,76,12))
trial_c_results = calculate_accuracy_stats_from_observations(obs,exp)
#Check that this is correct
self.assertFloatEqual(trial_c_results["false_positive_rate"],0.88)
trial_c = (obs,exp)
trials = [trial_a, trial_b,trial_c]
#Finally the actual test
obs_points = roc_points(trials)
exp_points = [(0.28,0.63),(0.77,0.77),(0.88,0.24)]
self.assertFloatEqual(obs_points,exp_points)
