def test_divide_by_scalar_samples() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0), rng=1234)
hist1.fill(np.random.normal(size=1000))
scale = 2.0
a1 = hist1.view() / scale
hist3 = hist1 / scale
assert np.array_equal(hist3.view(), a1)
def test_samples() -> None:
numsamples = 100
hist = BootstrapHistogram(
bh.axis.Regular(10, 0.0, 1.0),
bh.axis.Regular(10, 0.0, 1.0),
numsamples=numsamples,
rng=1234,
)
size = 100000
xdata = np.random.uniform(size=size)
ydata = np.random.uniform(size=size)
hist.fill(xdata, ydata)
XY = hist.view()
mean = np.average(XY, axis=2)
std = np.std(XY, axis=2)
nbins = len(hist.axes[0]) * len(hist.axes[1])
assert array_almost_equal(mean,
size / nbins,
delta=5.0 * np.sqrt(size / nbins))
assert array_almost_equal(
std,
np.sqrt(size / nbins),
delta=5.0 *
_standard_error_std(size=numsamples, sigma=np.sqrt(size / nbins)),
)
return
def test_view_property() -> None:
numsamples = 10
nbins = 5
hist = BootstrapHistogram(bh.axis.Regular(nbins, -5.0, 5.0),
numsamples=numsamples,
rng=1234)
view = hist.view()
assert np.array_equal(view, np.zeros(shape=(nbins, numsamples)))
def test_add_scalar() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(2, 0.0, 2.0), rng=1234)
hist1.fill([0.0, 0.0, 1.0])
hist3 = hist1 + 2
hist4 = 2 + hist1 # test radd method
assert np.array_equal(hist3.nominal.view(), [4.0, 3.0])
assert np.array_equal(hist3.samples, hist1.samples + 2)
assert hist3 == hist4
def test_divide_by_scalar_nominal() -> None:
hist = BootstrapHistogram(bh.axis.Regular(1, -1.0, 1.0),
numsamples=10,
rng=1234)
hist.fill(0.0)
hist.fill(0.0)
scaled = hist / 2.0
assert np.array_equal(list(hist.nominal.view()), [2.0])
assert np.array_equal(list(scaled.nominal.view()), [1.0])
def test_add() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0), rng=1234)
hist2 = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0), rng=1234)
hist1.fill(np.random.normal(size=1000))
hist2.fill(np.random.normal(size=1000))
a1 = hist1.view()
a2 = hist2.view()
hist3 = hist1 + hist2
assert np.array_equal(hist3.view(), a1 + a2)
def test_multiply_by_scalar_nominal() -> None:
hist = BootstrapHistogram(bh.axis.Regular(1, -1.0, 1.0),
numsamples=10,
rng=1234)
hist.fill(0.0)
scaled = hist * 2.0
scaled2 = 2.0 * hist # test rmul
assert np.array_equal(list(hist.nominal.view()), [1.0])
assert np.array_equal(list(scaled.nominal.view()), [2.0])
assert scaled2 == scaled
def test_fill_with_various_length_arrays(numsamples: int, numarray: int,
withseed: bool,
withweight: bool) -> None:
hist = BootstrapHistogram(bh.axis.Regular(5, 0.0, 1.0),
numsamples=numsamples,
rng=1234)
values = np.random.uniform(size=numarray)
weight = np.random.uniform(size=numarray) if withweight else None
seed = np.arange(numarray) if withseed else None
hist.fill(values, weight=weight, seed=seed)
def uniformhist(
ax: axis.Axis = axis.Variable([0.0, 0.1, 0.3, 0.6, 1.0]),
size: int = 100,
numsamples: int = 10,
low: float = 0.0,
high: float = 1.0,
) -> BootstrapHistogram:
hist = BootstrapHistogram(ax, numsamples=numsamples)
hist.fill(np.random.uniform(low=0.0, high=1.0, size=size))
return hist
def test_mean() -> None:
size = 100000
hist = BootstrapHistogram(bh.axis.Regular(100, 0.0, 1.0),
numsamples=100,
rng=1234)
data = np.random.uniform(size=size)
hist.fill(data)
nbins = len(hist.axes[0])
assert array_almost_equal(hist.mean(),
size / nbins,
delta=5.0 * np.sqrt(size / nbins))
return
def test_inequality() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0))
hist2 = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0))
data = np.random.normal(size=1000)
hist1.fill(data)
hist2.fill(data)
assert hist1 != hist2
def stddev_bootstrap_2d(nevents=100, numbins=1, numsamples=1000):
# create a 2D histogram
hist2d = BootstrapHistogram(
axis.Regular(numbins, 0.0, 1.0),
axis.Regular(numbins, 0.0, 1.0),
numsamples=numsamples,
)
# fill with some random data
hist2d.fill(np.random.uniform(size=nevents),
np.random.uniform(size=nevents))
histX = hist2d.project(0)
histY = hist2d.project(1)
histSum = histX + histY
return histSum.std()[0]
def test_std() -> None:
numsamples = 100
hist = BootstrapHistogram(bh.axis.Regular(100, 0.0, 1.0),
numsamples=numsamples,
rng=1234)
size = 100000
data = np.random.uniform(size=size)
hist.fill(data)
nbins = len(hist.axes[0])
assert array_almost_equal(
hist.std(),
np.sqrt(size / nbins),
delta=5.0 *
_standard_error_std(size=numsamples, sigma=np.sqrt(size / nbins)),
)
def test_divide_by_histogram() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(2, 0.0, 2.0), rng=1234)
hist2 = BootstrapHistogram(bh.axis.Regular(2, 0.0, 2.0), rng=1234)
hist1.fill([0.0, 1.0] * 400)
hist2.fill([0.0, 1.0] * 200)
hist3 = hist1 / hist2
assert np.array_equal(hist3.nominal.view(), [2.0, 2.0])
assert np.array_equal(hist3.samples, (hist1.samples / hist2.samples))
def test_sub_by_histogram() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(2, 0.0, 2.0), rng=1234)
hist2 = BootstrapHistogram(bh.axis.Regular(2, 0.0, 2.0), rng=1234)
hist1.fill([0.0, 1.0, 0.0, 1.0])
hist2.fill([0.0, 1.0, 0.0, 1.0, 0.0, 1.0])
hist3 = hist1 - hist2
assert np.array_equal(hist3.nominal.view(), [-1.0, -1.0])
assert np.array_equal(hist3.samples, hist1.samples - hist2.samples)
def uniformhist(
ax: Tuple[axis.Axis, axis.Axis] = (
axis.Regular(10, 0.0, 1.0),
axis.Regular(10, 0.0, 1.0),
),
size: int = 100,
numsamples: int = 10,
low: float = 0.0,
high: float = 1.0,
) -> BootstrapHistogram:
hist = BootstrapHistogram(*ax, numsamples=numsamples)
hist.fill(
np.random.uniform(low=0.0, high=1.0, size=size),
np.random.uniform(low=0.0, high=1.0, size=size),
)
return hist
def test_fill_with_record_id_seed() -> None:
hist1 = BootstrapHistogram(bh.axis.Regular(100, 0.0, 1.0),
numsamples=100,
rng=1234)
hist2 = BootstrapHistogram(bh.axis.Regular(100, 0.0, 1.0),
numsamples=100,
rng=5678)
data = np.random.uniform(size=10000)
seed = np.arange(len(data))
hist1.fill(data, seed=seed)
hist2.fill(data, seed=seed)
assert hist1 == hist2
def _run(
withweight: bool,
withseed: bool,
numsamples: int = 10000,
arraysize: int = 100,
numfills: int = 5,
fast: bool = True,
) -> float:
hist = BootstrapHistogram(bh.axis.Regular(100, 0.0, 1.0),
numsamples=numsamples,
rng=1234)
data = np.random.uniform(size=arraysize)
weight = np.random.uniform(size=arraysize) if withweight else None
seed = np.arange(arraysize) if withseed else None
if fast:
return timeit(lambda: hist._fill_fast(data, weight=weight, seed=seed),
number=numfills)
else:
return timeit(lambda: hist._fill_slow(data, weight=weight, seed=seed),
number=numfills)
def test_fill() -> None:
hist = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0),
numsamples=10,
rng=1234)
size = 100000
data = np.random.normal(loc=0.0, scale=1.0, size=size)
hist.fill(data)
x = hist.axes[0].centers
y = hist.view()[:, np.random.randint(0, hist.numsamples)]
mean = np.average(x, weights=y)
std = np.average((x - mean)**2, weights=y)
binwidth = hist.axes[0].edges[1] - hist.axes[0].edges[0]
assert array_almost_equal(mean,
0.0,
delta=5.0 * _standard_error_mean(size=size) +
binwidth)
assert array_almost_equal(std,
1.0,
delta=5.0 * _standard_error_std(size=size) +
binwidth)
return
def test_boostraphistogram_fast_and_slow_filling_give_identical_results(
withweight: bool, withseed: bool
) -> None:
size = 10000
hist_slow = BootstrapHistogram(
bh.axis.Regular(100, 0.0, 1.0), numsamples=100, rng=1234
)
hist_fast = BootstrapHistogram(
bh.axis.Regular(100, 0.0, 1.0), numsamples=100, rng=1234
)
data = np.random.uniform(size=size)
weight = np.random.uniform(size=size) if withweight else None
seed = np.arange(size) if withseed else None
hist_slow._fill_slow(data, weight=weight, seed=seed)
hist_fast._fill_fast(data, weight=weight, seed=seed)
assert hist_slow == hist_fast
def test_projection3() -> None:
hist = BootstrapHistogram(
bh.axis.Regular(3, 0.0, 3.0),
bh.axis.Regular(2, 0.0, 2.0),
numsamples=1000,
rng=1234,
)
X = [0.0, 1.0, 1.0, 2.0, 2.0, 2.0]
Y = [0.0, 1.0, 0.0, 1.0, 0.0, 0.0]
hist.fill(X, Y)
hx = hist.project(0)
hy = hist.project(1)
assert np.array_equal(hx.nominal.view(), np.array([1.0, 2.0, 3.0]))
assert np.array_equal(hy.nominal.view(), np.array([4.0, 2.0]))
delta = 0.1
assert array_almost_equal(
np.average(hx.samples.view(), axis=1),
np.array([1.0, 2.0, 3.0]),
delta=delta,
)
assert array_almost_equal(np.average(hy.samples.view(), axis=1),
np.array([4.0, 2.0]),
delta=delta)
def test_project() -> None:
xax = bh.axis.Regular(2, 0.0, 2.0)
yax = bh.axis.Regular(3, 0.0, 3.0)
hist = BootstrapHistogram(xax, yax, numsamples=9, rng=1234)
assert hist.view(flow=True).shape == (2 + 2, 3 + 2, 9)
assert hist.project(0).view(flow=True).shape == (2 + 2, 9)
assert hist.project(1).view(flow=True).shape == (3 + 2, 9)
assert hist.project(2).view(flow=True).shape == (9, )
assert hist.project(0, 1).view(flow=True).shape == (2 + 2, 3 + 2, 9)
def stddev_bootstrap_1d_eventID(nevents=100, numbins=1, numsamples=1000):
# create a 2D histogram
histX = BootstrapHistogram(axis.Regular(numbins, 0.0, 1.0),
numsamples=numsamples)
histY = BootstrapHistogram(axis.Regular(numbins, 0.0, 1.0),
numsamples=numsamples)
# fill with some random data
eventID = np.arange(nevents)
X = np.random.uniform(size=nevents)
Y = np.random.uniform(size=nevents)
histX = histX.fill(X, seed=eventID)
histY = histY.fill(Y, seed=eventID)
histSum = histX + histY
return histSum.std()[0]
def stddev_bootstrap_1d(nevents=100, numbins=1, numsamples=1000):
# create a 2D histogram
histX = BootstrapHistogram(axis.Regular(numbins, 0.0, 1.0),
numsamples=numsamples,
rng=1234)
histY = BootstrapHistogram(axis.Regular(numbins, 0.0, 1.0),
numsamples=numsamples,
rng=1234)
# fill with some random data
X = np.random.uniform(size=nevents)
Y = np.random.uniform(size=nevents)
histX = histX.fill(X)
histY = histY.fill(Y)
histSum = histX + histY
return histSum.std()[0]
def test_boostraphistogram_fast_is_faster_than_slow(
withweight: bool, withseed: bool
) -> None:
numsamples = 10000
arraysize = 100
hist_slow = BootstrapHistogram(
bh.axis.Regular(100, 0.0, 1.0), numsamples=numsamples, rng=1234
)
hist_fast = BootstrapHistogram(
bh.axis.Regular(100, 0.0, 1.0), numsamples=numsamples, rng=1234
)
data = np.random.uniform(size=arraysize)
weight = np.random.uniform(size=arraysize) if withweight else None
seed = np.arange(arraysize) if withseed else None
slowtime = timeit(
lambda: hist_slow._fill_slow(data, weight=weight, seed=seed), number=5
)
fasttime = timeit(
lambda: hist_fast._fill_fast(data, weight=weight, seed=seed), number=5
)
assert slowtime > fasttime
def test_fill_with_integer_weights() -> None:
hist = BootstrapHistogram(bh.axis.Regular(5, 0.0, 5.0), rng=1234)
values = [0.0, 1.0, 2.0, 3.0, 3.0, 4.0]
weights = [0, 1, 2, 3, 4, -1]
hist.fill(values, weight=weights)
assert np.array_equal(hist.nominal.view(), [0.0, 1.0, 2.0, 7.0, -1.0])
def test_fill_with_invalid_data() -> None:
hist1d = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0),
numsamples=10,
rng=1234)
hist2d = BootstrapHistogram(
bh.axis.Regular(100, -5.0, 5.0),
bh.axis.Regular(100, -5.0, 5.0),
numsamples=10,
rng=1234,
)
with pytest.raises(ValueError):
hist1d.fill()
with pytest.raises(ValueError):
hist2d.fill([1, 2], [1])
with pytest.raises(ValueError):
hist2d.fill([1], [1, 2])
with pytest.raises(ValueError):
hist1d.fill([1, 2], weight=[1])
with pytest.raises(ValueError):
hist1d.fill([1], weight=[1, 2])
with pytest.raises(ValueError):
hist1d.fill([1], seed=[1, 2])
with pytest.raises(ValueError):
hist1d.fill([1, 2], seed=[1])
return
def test_fill_with_empty_weighted_array() -> None:
hist = BootstrapHistogram(bh.axis.Regular(2, 0.0, 2.0),
numsamples=9,
rng=1234)
hist.fill([], weight=[])
assert np.all(hist.view(flow=True) == np.zeros((2 + 2, 9)))
def test_numsamples_property() -> None:
numsamples = 100
hist = BootstrapHistogram(bh.axis.Regular(100, -5.0, 5.0),
numsamples=numsamples,
rng=1234)
assert hist.numsamples == numsamples
def test_axes_property() -> None:
axes = (bh.axis.Regular(100, -5.0, 5.0), )
hist = BootstrapHistogram(*axes, rng=1234)
assert hist.axes[:-1] == axes
