def plot_heated_stacked_area(df: pd.DataFrame,
lines: str,
heat: str,
backtest: str = None,
reset_y_lim: bool = False,
figsize: Tuple[int, int] = (16, 9),
color_map: str = 'afmhot',
ax: Axis = None,
upper_lower_missing_scale: float = 0.05) -> Axis:
color_function = plt.get_cmap(color_map)
x = df.index
y = df[lines].values
c = df[heat].values
b = df[backtest].values if backtest is not None else None
# assert enough data
assert len(y.shape) > 1 and len(
c.shape) > 1, "lines and heat need to be 2 dimensions!"
# make sure we have one more line as heats
if c.shape[1] == y.shape[1] + 1:
lower = np.full((c.shape[0], 1),
y.min() * (1 - upper_lower_missing_scale))
upper = np.full((c.shape[0], 1),
y.max() * (1 + upper_lower_missing_scale))
y = np.hstack([lower, y, upper])
# check for matching columns
assert y.shape[1] - 1 == c.shape[
1], f'unexpeced shapes: {y.shape[1] - 1} != {c.shape[1]}'
_, ax = plt.subplots(figsize=figsize) if ax is None else (None, ax)
ax.plot(x, y, color='k', alpha=0.0)
for ci in range(c.shape[1]):
for xi in range(len(x)):
ax.fill_between(x[xi - 1:xi + 1],
y[xi - 1:xi + 1, ci],
y[xi - 1:xi + 1, ci + 1],
facecolors=color_function(c[xi - 1:xi + 1, ci]))
if ci > 0:
# todo annotate all first last and only convert date if it is actually a date
ax.annotate(f'{y[-1, ci]:.2f}',
xy=(mdates.date2num(x[-1]), y[-1, ci]),
xytext=(4, -4),
textcoords='offset pixels')
# reset limits
ax.autoscale(tight=True)
if reset_y_lim:
ax.set_ylim(bottom=y[:, 1].min(), top=y[:, -1].max())
# backtest
if backtest:
ax.plot(x, b)
return ax
def ppk_plot(data: (List[int], List[float], pd.Series, np.array),
upper_control_limit: (int, float),
lower_control_limit: (int, float),
threshold_percent: float = 0.001,
ax: Axis = None):
"""
Shows the statistical distribution of the data along with CPK and limits.
:param data: a list, pandas.Series, or numpy.array representing the data set
:param upper_control_limit: an integer or float which represents the upper control limit, commonly called the UCL
:param lower_control_limit: an integer or float which represents the upper control limit, commonly called the UCL
:param threshold_percent: the threshold at which % of units above/below the number will display on the plot
:param ax: an instance of matplotlig.axis.Axis
:return: None
"""
data = coerce(data)
mean = data.mean()
std = data.std()
if ax is None:
fig, ax = plt.subplots()
ax.hist(data, density=True, label='data', alpha=0.3)
x = np.linspace(mean - 4 * std, mean + 4 * std, 100)
pdf = stats.norm.pdf(x, mean, std)
ax.plot(x, pdf, label='normal fit', alpha=0.7)
bottom, top = ax.get_ylim()
ax.axvline(mean, linestyle='--')
ax.text(mean, top * 1.01, s='$\mu$', ha='center')
ax.axvline(mean + std, alpha=0.6, linestyle='--')
ax.text(mean + std, top * 1.01, s='$\sigma$', ha='center')
ax.axvline(mean - std, alpha=0.6, linestyle='--')
ax.text(mean - std, top * 1.01, s='$-\sigma$', ha='center')
ax.axvline(mean + 2 * std, alpha=0.4, linestyle='--')
ax.text(mean + 2 * std, top * 1.01, s='$2\sigma$', ha='center')
ax.axvline(mean - 2 * std, alpha=0.4, linestyle='--')
ax.text(mean - 2 * std, top * 1.01, s='-$2\sigma$', ha='center')
ax.axvline(mean + 3 * std, alpha=0.2, linestyle='--')
ax.text(mean + 3 * std, top * 1.01, s='$3\sigma$', ha='center')
ax.axvline(mean - 3 * std, alpha=0.2, linestyle='--')
ax.text(mean - 3 * std, top * 1.01, s='-$3\sigma$', ha='center')
ax.fill_between(x,
pdf,
where=x < lower_control_limit,
facecolor='red',
alpha=0.5)
ax.fill_between(x,
pdf,
where=x > upper_control_limit,
facecolor='red',
alpha=0.5)
lower_percent = 100.0 * stats.norm.cdf(lower_control_limit, mean, std)
lower_percent_text = f'{lower_percent:.02f}% < LCL' if lower_percent > threshold_percent else None
higher_percent = 100.0 - 100.0 * stats.norm.cdf(upper_control_limit, mean,
std)
higher_percent_text = f'{higher_percent:.02f}% > UCL' if higher_percent > threshold_percent else None
left, right = ax.get_xlim()
bottom, top = ax.get_ylim()
cpk = calc_ppk(data,
upper_control_limit=upper_control_limit,
lower_control_limit=lower_control_limit)
lower_sigma_level = (mean - lower_control_limit) / std
if lower_sigma_level < 6.0:
ax.axvline(lower_control_limit,
color='red',
alpha=0.25,
label='limits')
ax.text(lower_control_limit,
top * 0.95,
s=f'$-{lower_sigma_level:.01f}\sigma$',
ha='center')
else:
ax.text(left, top * 0.95, s=f'limit > $-6\sigma$', ha='left')
upper_sigma_level = (upper_control_limit - mean) / std
if upper_sigma_level < 6.0:
ax.axvline(upper_control_limit, color='red', alpha=0.25)
ax.text(upper_control_limit,
top * 0.95,
s=f'${upper_sigma_level:.01f}\sigma$',
ha='center')
else:
ax.text(right, top * 0.95, s=f'limit > $6\sigma$', ha='right')
strings = [f'Ppk = {cpk:.02f}']
strings.append(f'$\mu = {mean:.3g}$')
strings.append(f'$\sigma = {std:.3g}$')
if lower_percent_text:
strings.append(lower_percent_text)
if higher_percent_text:
strings.append(higher_percent_text)
props = dict(boxstyle='round',
facecolor='white',
alpha=0.75,
edgecolor='grey')
ax.text(right - (right - left) * 0.05,
0.85 * top,
'\n'.join(strings),
bbox=props,
ha='right',
va='top')
ax.legend(loc='lower right')