def test_with_swap_rate():
dm = DataMatrix(length=N)
for precision, guess_rate, swap_rate in itertools.product(
(500, 2000), (0, .25), (0, .25)):
dm.target = np.random.randint(0, 359, N)
dm.nontarget1 = dm.target + 180
dm.responses = dm.target[:]
n_guess = int(N * guess_rate)
n_swap = int(N * swap_rate)
dm.responses[:n_guess] = np.random.randint(0, 359, n_guess)
dm.responses[n_guess:n_guess +
n_swap] = dm.nontarget1[n_guess:n_guess + n_swap]
dm.responses += np.random.normal(loc=0,
scale=_precision_to_scale(precision),
size=N)
p, gr, sr = bmt.fit_mixture_model(
x=bmt.response_bias(dm.target, dm.responses),
x_nontargets=[bmt.response_bias(dm.nontarget1, dm.responses)],
include_bias=False)
assert (math.isclose(precision, p, rel_tol=.25))
assert (math.isclose(guess_rate, gr, abs_tol=2))
assert (math.isclose(swap_rate, sr, abs_tol=.1))
def test_with_bias_and_swap_rate():
dm = DataMatrix(length=N)
for precision, guess_rate, bias, swap_rate in itertools.product(
(500, 2000), (0, .25), (0, 2.5), (0, .25)):
dm.target = np.random.randint(0, 359, N)
dm.nontarget1 = dm.target + 180
dm.responses = dm.target[:]
n_guess = int(N * guess_rate)
n_swap = int(N * swap_rate)
dm.responses[:n_guess] = np.random.randint(0, 359, n_guess)
dm.responses[n_guess:n_guess +
n_swap] = dm.nontarget1[n_guess:n_guess + n_swap]
dm.responses += np.random.normal(loc=0,
scale=_precision_to_scale(precision),
size=N)
for row in dm:
category = bmt.category(row.target, bmt.DEFAULT_CATEGORIES)
lower, upper, proto = bmt.DEFAULT_CATEGORIES[category]
if bmt._distance(row.responses, proto) > 0:
row.responses += bias
else:
row.responses -= bias
p, gr, b, sr = bmt.fit_mixture_model(
x=bmt.response_bias(dm.target,
dm.responses,
categories=bmt.DEFAULT_CATEGORIES),
x_nontargets=[
bmt.response_bias(dm.nontarget1,
dm.responses,
categories=bmt.DEFAULT_CATEGORIES)
])
assert (math.isclose(precision, p, rel_tol=.25))
assert (math.isclose(guess_rate, gr, abs_tol=.1))
assert (math.isclose(bias, b, abs_tol=2))
assert (math.isclose(swap_rate, sr, abs_tol=.1))
