def test_numpyAPI_var1d():
x = np.random.randn(5000)
bins = [-1.2, -1, -0.2, 0.7, 1.5, 2.1]
w = np.random.uniform(0.5, 1.9, 5000)
pygram_h, __ = pygram11.histogram(x, bins=bins)
numpy_h, __ = np.histogram(x, bins=bins)
npt.assert_almost_equal(pygram_h, numpy_h, 5)
pygram_h, __ = pygram11.histogram(x, bins=bins, weights=w)
numpy_h, __ = np.histogram(x, bins=bins, weights=w)
npt.assert_almost_equal(pygram_h, numpy_h, 5)
def test_numpyAPI_fix1d():
x = np.random.randn(5000)
bins = 25
w = np.random.uniform(0.8, 1, 5000)
pygram_h, __ = pygram11.histogram(x, bins=25, range=(-3, 3))
numpy_h, __ = np.histogram(x, bins=np.linspace(-3, 3, 26))
npt.assert_almost_equal(pygram_h, numpy_h, 5)
pygram_h, __ = pygram11.histogram(x, bins=25, range=(-3, 3), weights=w)
numpy_h, __ = np.histogram(x, bins=np.linspace(-3, 3, 26), weights=w)
npt.assert_almost_equal(pygram_h, numpy_h, 5)
def dist_comparison_plot(
var: str,
region: str,
binning: tuple[int, float, float],
df: pd.DataFrame,
y: np.ndarray,
w: np.ndarray,
meta_table,
outdir: Path,
) -> None:
"""Compare shapes of BDT variable."""
is_tW = y == 1
is_tt = y == 0
w_tW = w[is_tW]
w_tt = w[is_tt]
tW_dist = df[var][is_tW].to_numpy()
tt_dist = df[var][is_tt].to_numpy()
bins = np.linspace(binning[1], binning[2], binning[0] + 1)
centers = bin_centers(bins)
bw = bins[1] - bins[0]
fig, (ax, axr) = plt.subplots(
2,
1,
sharex=True,
gridspec_kw=dict(height_ratios=[3.25, 1], hspace=0.025),
)
n1, e1 = histogram(tW_dist, weights=w_tW, bins=bins, density=True, flow=True)
n2, e2 = histogram(tt_dist, weights=w_tt, bins=bins, density=True, flow=True)
r = n1 / n2
ax.hist(centers, weights=n1, bins=bins, label=r"$tW$", histtype="step")
ax.hist(centers, weights=n2, bins=bins, label=r"$t\bar{t}$", histtype="step")
axr.hist(centers, bins=bins, weights=r, histtype="step", color="black")
ax.set_ylabel("Arb. Units", ha="right", y=1.0)
xl, yl = meta_axis_label(var, bw, meta_table)
axr.set_xlabel(xl, ha="right", x=1.0)
axr.set_xlim([binning[1], binning[2]])
axr.set_ylabel(r"$tW/t\bar{t}$")
ax.set_ylim([ax.get_ylim()[0], ax.get_ylim()[1] * 1.35])
ax.legend()
axr.axhline(y=1, ls="--", color="gray")
if np.amax(r) > 2:
axr.set_ylim([0, 3])
else:
axr.set_ylim([0, 2])
tdub.art.draw_atlas_label(ax, thesis=True)
fig.savefig(outdir / f"r{region}_SC_{var}.pdf")
plt.close(fig)
def test_flow_omp_var():
x = np.random.randn(100000)
bins = [-2, -1.7, -0.5, 0.2, 2.2]
pygram_h, __ = pygram11.histogram(x, bins=bins, omp=True, flow=True)
numpy_h, __ = np.histogram(x, bins=bins)
numpy_h[0] += sum(x < bins[0])
numpy_h[-1] += sum(x > bins[-1])
assert np.all(pygram_h == numpy_h)
def test_flow_omp():
x = np.random.randn(100000)
nbins = 50
rg = (-3, 3)
pygram_h, __ = pygram11.histogram(x,
bins=nbins,
range=rg,
omp=True,
flow=True)
numpy_h, __ = np.histogram(x, bins=nbins, range=rg)
numpy_h[0] += sum(x < rg[0])
numpy_h[-1] += sum(x > rg[1])
assert np.all(pygram_h == numpy_h)
def test_flow_weights_omp_var():
x = np.random.randn(100000)
w = np.random.uniform(0.5, 0.8, x.shape[0])
bins = [-2, -1.7, -0.5, 0.2, 2.2]
pygram_h, __ = pygram11.histogram(x,
bins=bins,
weights=w,
omp=True,
flow=True)
numpy_h, __ = np.histogram(x, bins=bins, weights=w)
numpy_h[0] += sum(w[x < bins[0]])
numpy_h[-1] += sum(w[x > bins[-1]])
assert np.allclose(pygram_h, numpy_h)
def test_flow_weights_omp():
x = np.random.randn(100000)
w = np.random.uniform(0.5, 0.8, x.shape[0])
nbins = 50
rg = (-3, 3)
pygram_h, __ = pygram11.histogram(x,
bins=nbins,
range=rg,
weights=w,
omp=True,
flow=True)
numpy_h, __ = np.histogram(x, bins=nbins, range=rg, weights=w)
numpy_h[0] += sum(w[x < rg[0]])
numpy_h[-1] += sum(w[x > rg[1]])
assert np.allclose(pygram_h, numpy_h)
def draw_stack(
*,
data_df: pd.DataFrame,
mc_dfs: List[pd.DataFrame],
distribution: str,
weight_name: str = "weight_nominal",
bins: Union[int, Sequence[numbers.Real]] = 10,
range: Optional[Tuple[float, float]] = None,
colors: Optional[Iterable[Any]] = None,
labels: Optional[Iterable[str]] = None,
lumi: float = 139.0,
legend_ncol: int = 2,
y_scalefac: float = 1.35,
) -> Tuple[plt.Figure, plt.Axes, plt.Axes]:
"""Given dataframes draw the stacked histograms for a distribution.
Parameters
----------
data_df : pandas.DataFrame
the dataframe for data
mc_dfs : list(pandas.DataFrame)
the list of MC dataframes
distribution: str
the variable to histogram
weight_name : str
the name of the weight column
bins : int or sequence of scalars
the number of bins or sequence representing bin edges
range : tuple(float, float), optional
the range to histogram the distribution (used for integral
bins, ignored if ``bins`` is a sequence).
colors : list(Any), optional
the colors for the Monte Carlo histograms, ``None`` defaults to
the normal colors associated with our standard samples
labels : list(str), optional
the list of labels for the legend. ``None`` default sto the the
normal labels associated with out standard samples
lumi : float
the luminosity for the data (to scale the MC)
legend_ncol : int
number of columns for the legend
y_scalefac : float
factor to scale the default maximum y-axis range by
Returns
-------
:py:obj:`matplotlib.figure.Figure`
the figure associated with the axes
:py:obj:`matplotlib.axes.Axes`
the main axis object which has the plot
:py:obj:`matplotlib.axes.Axes`
the axis object which has the ratio plot
Examples
--------
>>> import matplotlib.pyplot as plt
>>> from tdub.raw_art import draw_stack
>>> mc_dfs = get_mc_dataframes() # user defined function returns a list of dataframes
>>> data_df = get_data_dataframe() # user defined function returns a single dataframe
>>> colors = list(reversed(["#1f77b4", "#d62728", "#2ca02c", "#ff7f0e", "#9467bd"]))
>>> labels = list(reversed(["$tW$", "$t\\bar{t}$", "Diboson", "$Z+$jets", "MCNP"]))
>>> fig, ax, axr = draw_stacks(
... data_df=datadf,
... mc_dfs=mc_dfs,
... labels=labels,
... colors=colors,
... distribution="mass_lep1jet2",
... bins=25,
... range=(0, 250.0),
... )
>>> fig.savefig("mass_lep1jet2.pdf")
>>> plt.close(fig)
"""
data_count, __ = pygram11.histogram(data_df[distribution].to_numpy(),
bins=bins,
range=range,
flow=True)
data_err = np.sqrt(data_count)
mc_dists = [df[distribution].to_numpy() for df in mc_dfs]
mc_ws = [df[weight_name].to_numpy() * lumi for df in mc_dfs]
mc_hists = [
pygram11.histogram(mcd, weights=mcw, bins=bins, range=range, flow=True)
for mcd, mcw in zip(mc_dists, mc_ws)
]
mc_counts = [mcc[0] for mcc in mc_hists]
mc_errs = [mcc[1] for mcc in mc_hists]
mc_total = np.sum(mc_counts, axis=0)
ratio = data_count / mc_total
mc_total_err = np.sqrt(np.sum([mce**2 for mce in mc_errs], axis=0))
ratio_err = data_count / (mc_total**2) + np.power(
data_count * mc_total_err / (mc_total**2), 2)
if colors is None:
colors = ["#1f77b4", "#d62728", "#2ca02c", "#ff7f0e", "#9467bd"]
colors.reverse()
if labels is None:
labels = ["$tW$", "$t\\bar{t}$", "Diboson", "$Z+$jets", "MCNP"]
labels.reverse()
edges, centers = edges_and_centers(bins, range=range)
fig, (ax, axr) = plt.subplots(
2,
1,
sharex=True,
figsize=(6, 5.25),
gridspec_kw=dict(height_ratios=[3.25, 1], hspace=0.025),
)
ax.hist(
[centers for _ in labels],
weights=mc_counts,
bins=edges,
histtype="stepfilled",
label=labels,
color=colors,
stacked=True,
)
ax.errorbar(centers,
data_count,
yerr=data_err,
fmt="ko",
label="Data",
zorder=999)
ax.set_ylim([0, ax.get_ylim()[1] * y_scalefac])
ax.legend(loc="upper right")
handles, labels = ax.get_legend_handles_labels()
handles.insert(0, handles.pop())
labels.insert(0, labels.pop())
ax.legend(handles, labels, loc="upper right", ncol=legend_ncol)
axr.errorbar(centers, ratio, yerr=ratio_err, fmt="ko", zorder=999)
axr.plot([edges[0], edges[-1]], [1, 1],
color="gray",
linestyle="solid",
marker=None)
axr.set_ylim([0.8, 1.2])
axr.set_yticks([0.9, 1.0, 1.1])
axr.autoscale(enable=True, axis="x", tight=True)
return fig, ax, axr
def plot_from_region_frames(
frames: Dict[str, pd.DataFrame],
variable: str,
binning: Tuple[int, float, float],
region_label: str,
logy: bool = False,
legend_kw: Dict[str, Any] = None,
) -> Tuple[plt.Figure, plt.Axes, plt.Axes]:
"""create a histogram plot pdf from dataframes and a desired variable
Parameters
----------
frames : dict(str, pd.DataFrame)
the dataframes for all samples
variable : str
the variable we want to histogram
binning : tuple(int, float, float)
the bin definition
region_label : str
the region label (will be part of out file name)
logy : bool
if true set the yscale to log
legend_kw : dict(str, Any)
keyward arguments passed to :py:func:`matplotlib.Axes.axes.legend`.
"""
if variable not in frames["Data"].columns.to_list():
log.warning("%s not in dataframe; skipping" % variable)
return None, None, None
nbins, start, stop = binning
bin_edges = np.linspace(start, stop, nbins + 1)
counts = {}
errors = {}
for samp in ALL_SAMPLES:
x = frames[samp][variable].to_numpy()
w = frames[samp]["weight_nominal"].to_numpy()
count, err = pg.histogram(x,
bins=nbins,
range=(start, stop),
weights=w,
flow=True)
counts[samp] = count
errors[samp] = err
fig, ax, axr = canvas_from_counts(counts, errors, bin_edges)
draw_atlas_label(
ax,
extra_lines=["$\sqrt{s} = 13$ TeV, $L = 139$ fb$^{-1}$", region_label])
tune_axes(ax, axr, variable, binning, logy=logy)
if legend_kw is None:
legend_kw = {}
legend_kw["ncol"] = 2
ax.legend(loc="upper right")
handles, labels = ax.get_legend_handles_labels()
handles.insert(0, handles.pop())
labels.insert(0, labels.pop())
ax.legend(handles, labels, loc="upper right", **legend_kw)
fig.subplots_adjust(left=0.125, bottom=0.095, right=0.965, top=0.95)
return fig, ax, axr
