def test_kendall_tau_all(): assert abs(ccfcr.kendall_tau(prev_data, curr_data) - kmc.kendall_tau_all(prev_data, curr_data, sorted_prev_data)[0]) < epsilon # original assert abs(ccfcr.kendall_tau_w(prev_data, curr_data) - kmc.kendall_tau_all(prev_data, curr_data, sorted_prev_data)[1]) < epsilon # weighted assert abs(ccfcr.kendall_tau(sorted_prev_data, sorted_prev_data) - kmc.kendall_tau_all(prev_data, curr_data, sorted_prev_data)[2]) < epsilon # original for list_a assert abs(ccfcr.kendall_tau_w(sorted_prev_data, sorted_prev_data) - kmc.kendall_tau_all(prev_data, curr_data, sorted_prev_data)[3]) < epsilon # weighted for list_a assert abs(ccfcr.kendall_tau(curr_data, curr_data) - kmc.kendall_tau_all(prev_data, curr_data, sorted_prev_data)[4]) < epsilon # original for list_b assert abs(ccfcr.kendall_tau_w(curr_data, curr_data) - kmc.kendall_tau_all(prev_data, curr_data, sorted_prev_data)[5]) < epsilon # weighted for list_b
def compute_kendall(list_a, list_b):
print(" [ kendall start ... ]")
val = ccfcr.kendall_tau(list_a, list_b)
w_val = ccfcr.kendall_tau_w(list_a, list_b)
sorted_a = sorted(list_a, reverse=False) # due to position ranking ascending order is needed
# list_b = sorted(list_b, reverse=False)
val_a = ccfcr.kendall_tau(sorted_a, sorted_a)
val_b = ccfcr.kendall_tau(list_b, list_b)
w_val_a = ccfcr.kendall_tau_w(sorted_a, sorted_a)
w_val_b = ccfcr.kendall_tau_w(list_b, list_b )
print(" [ kendall done ]")
return [val/math.sqrt(val_a*val_b), w_val/ math.sqrt(w_val_a*w_val_b)]
def test_kendall_tau_w(): assert abs(ccfcr.kendall_tau_w(prev_data_with_centrality, curr_data_with_centrality) - ccfcr.kendall_tau_w(prev_data_with_position, curr_data_with_position)) < epsilon assert False == (abs(ccfcr.kendall_tau_w(prev_data_with_centrality, prev_data_with_centrality) - ccfcr.kendall_tau_w(prev_data_with_position, curr_data_with_position)) < epsilon)
