def test_qth_survival_time_with_cdf_instead_of_survival_function():
cdf = np.linspace(0, 1, 50)
assert utils.qth_survival_times(0.5, cdf, cdf=True) == 25
assert utils.qth_survival_times(0.05, cdf, cdf=True) == 3
cdf = np.linspace(0.1, 1, 50)
assert utils.qth_survival_times(0.05, cdf, cdf=True) == -np.inf
assert utils.qth_survival_times(0.50, cdf, cdf=True) == 22
def test_qth_survival_times_with_multivariate_q():
sf = np.linspace(1, 0, 50)
sf_multi_df = pd.DataFrame({'sf': sf, 'sf**2': sf ** 2})
assert_frame_equal(utils.qth_survival_times([0.2, 0.5], sf_multi_df), pd.DataFrame([[40, 25], [28, 15]], columns=[0.2, 0.5], index=['sf', 'sf**2']))
assert_frame_equal(utils.qth_survival_times([0.2, 0.5], sf_multi_df['sf']), pd.DataFrame([[40, 25]], columns=[0.2, 0.5], index=['sf']))
assert_frame_equal(utils.qth_survival_times(0.5, sf_multi_df), pd.DataFrame([[25], [15]], columns=[0.5], index=['sf', 'sf**2']))
assert utils.qth_survival_times(0.5, sf_multi_df['sf']) == 25
def test_qth_survival_times_with_multivariate_q():
sf = np.linspace(1, 0, 50)
sf_multi_df = pd.DataFrame({'sf': sf, 'sf**2': sf ** 2})
assert_frame_equal(utils.qth_survival_times([0.2, 0.5], sf_multi_df), pd.DataFrame([[40, 28], [25, 15]], index=[0.2, 0.5], columns=['sf', 'sf**2']))
assert_frame_equal(utils.qth_survival_times([0.2, 0.5], sf_multi_df['sf']), pd.DataFrame([40, 25], index=[0.2, 0.5], columns=['sf']))
assert_frame_equal(utils.qth_survival_times(0.5, sf_multi_df), pd.DataFrame([[25, 15]], index=[0.5], columns=['sf', 'sf**2']))
assert utils.qth_survival_times(0.5, sf_multi_df['sf']) == 25
def test_qth_survival_times_with_duplicate_q_returns_valid_index_and_shape():
sf = pd.DataFrame(np.linspace(1, 0, 50))
q = pd.Series([0.5, 0.5, 0.2, 0.0, 0.0])
actual = utils.qth_survival_times(q, sf)
assert actual.shape[0] == len(q)
npt.assert_almost_equal(actual.index.values, q.values)
def predict_percentile(self, X, p=0.5):
"""
Returns the median lifetimes for the individuals, by default. If the survival curve of an
individual does not cross 0.5 in the timeline (set in ``fit``), then the result is infinity.
http://stats.stackexchange.com/questions/102986/percentile-loss-functions
Parameters
----------
X: numpy array or DataFrame
a (n,d) covariate numpy array or DataFrame. If a DataFrame, columns
can be in any order. If a numpy array, columns must be in the
same order as the training data.
p: float, optional (default=0.5)
the percentile, must be between 0 and 1.
Returns
-------
percentiles: DataFrame
See Also
--------
predict_median
"""
subjects = _get_index(X)
return qth_survival_times(p, self.predict_survival_function(X)[subjects]).T
def _conditional_time_to_event_(self):
"""
Return a DataFrame, with index equal to survival_function_, that estimates the median
duration remaining until the death event, given survival up until time t. For example, if an
individual exists until age 1, their expected life remaining *given they lived to time 1*
might be 9 years.
Returns
-------
conditional_time_to_: DataFrame
with index equal to survival_function_
"""
age = self.survival_function_.index.values[:, None]
columns = ["%s - Conditional time remaining to event" % self._label]
return (
pd.DataFrame(
qth_survival_times(self.survival_function_[self._label] * 0.5, self.survival_function_)
.sort_index(ascending=False)
.values,
index=self.survival_function_.index,
columns=columns,
)
- age
)
def predict_percentile(self, X, p=0.5):
"""
X: a (n,d) covariate matrix
Returns the median lifetimes for the individuals.
http://stats.stackexchange.com/questions/102986/percentile-loss-functions
"""
index = _get_index(X)
return qth_survival_times(p, self.predict_survival_function(X)[index])
def predict_percentile(self, X, p=0.5):
"""
X: a (n,d) covariate matrix
Returns the median lifetimes for the individuals.
http://stats.stackexchange.com/questions/102986/percentile-loss-functions
"""
index = _get_index(X)
return qth_survival_times(p, self.predict_survival_function(X)[index])
def predict_percentile(self, conditioned_sf, percentile):
# Predict the month number where the survival chance of customer is 50%
# This can also be modified as predictions_50 = qth_survival_times(.50, conditioned_sf),
# where the percentile can be modified depending on our requirement
predictions = qth_survival_times(percentile, conditioned_sf)
st.write(
'### predictions\n Predicting the month at which the survival chance of the customer is ',
percentile * 100, ' percentile')
st.write(predictions[[self.customer]])
return predictions
def predict_percentile(self, X, p=0.5):
"""
X: a (n,d) covariate numpy array or DataFrame. If a DataFrame, columns
can be in any order. If a numpy array, columns must be in the
same order as the training data.
By default, returns the median lifetimes for the individuals.
http://stats.stackexchange.com/questions/102986/percentile-loss-functions
"""
index = _get_index(X)
return qth_survival_times(p, self.predict_survival_function(X)[index])
def predict_percentile(self, X, p=0.5):
"""
X: a (n,d) covariate numpy array or DataFrame. If a DataFrame, columns
can be in any order. If a numpy array, columns must be in the
same order as the training data.
By default, returns the median lifetimes for the individuals.
http://stats.stackexchange.com/questions/102986/percentile-loss-functions
"""
index = _get_index(X)
return qth_survival_times(p, self.predict_survival_function(X)[index])
def predict(self, df, q=0.5):
df, _ = self._prepare_df(df)
if self.scaler is not None:
X = df[self.feat_cols].values
X_scaled = self.scaler.transform(X)
df[self.feat_cols] = X_scaled
if self.verbose:
print("Scaled features based on training set results!")
# the cox model predicts log hazards and from that can also compute individual survival time curves
surv_funs = self.model.predict_survival_function(df)
# and we take the time point were the curve reaches 0.5
pred_time = qth_survival_times(q, surv_funs).squeeze()
# print("head(df) =\n{}".format(df.head(5)))
log_hazards = self.model.predict_log_partial_hazard(
df).values.squeeze()
# print("Log_hazards = {}".format(log_hazards.shape))
# print("pred_time.shape", pred_time.shape)
# print("ids.shape", df[self.id_col].values.shape)
pred_df = pd.DataFrame({
self.id_col: df.index.values,
'pred_time': pred_time,
'pred_per_pat(log_hazard)': log_hazards
})
perf_df = None
# in case we have labels
if self.time_col in df.columns and self.event_col in df.columns:
# we can append some information to the pred_df as well
true_time = df[self.time_col].values
event_status = df[self.event_col].values
diff = true_time - pred_time
pred_df[self.event_col + "_truth"] = event_status
pred_df[self.time_col + "_truth"] = true_time
pred_df["error(time_prediction)"] = diff
# only if we have the true label we can return performance
# values
perf_df = pd.DataFrame({
'MSE': [np.mean(diff**2)],
'MAE': [np.mean(np.abs(diff))],
'C-index_time':
[concordance_index(true_time, pred_time, event_status)],
'C-index_log_hazard':
[concordance_index(true_time, log_hazards, event_status)]
})
return pred_df, perf_df
def predict_percentile(self, X, p=0.5):
"""
X: a (n,d) covariate numpy array or DataFrame. If a DataFrame, columns
can be in any order. If a numpy array, columns must be in the
same order as the training data.
Returns the median lifetimes for the individuals, by default. If the survival curve of an
individual does not cross 0.5, then the result is infinity.
http://stats.stackexchange.com/questions/102986/percentile-loss-functions
"""
subjects = _get_index(X)
return qth_survival_times(
p,
self.predict_survival_function(X)[subjects]).T
def _conditional_time_to_event_(self):
"""
Return a DataFrame, with index equal to survival_function_, that estimates the median
duration remaining until the death event, given survival up until time t. For example, if an
individual exists until age 1, their expected life remaining *given they lived to time 1*
might be 9 years.
Returns:
conditional_time_to_: DataFrame, with index equal to survival_function_
"""
age = self.survival_function_.index.values[:, None]
columns = ['%s - Conditional time remaining to event' % self._label]
return pd.DataFrame(qth_survival_times(self.survival_function_[self._label] * 0.5, self.survival_function_).T.sort_index(ascending=False).values,
index=self.survival_function_.index,
columns=columns) - age
def test_qth_survival_times_with_varying_datatype_inputs():
sf_list = [1.0, 0.75, 0.5, 0.25, 0.0]
sf_array = np.array([1.0, 0.75, 0.5, 0.25, 0.0])
sf_df_no_index = pd.DataFrame([1.0, 0.75, 0.5, 0.25, 0.0])
sf_df_index = pd.DataFrame([1.0, 0.75, 0.5, 0.25, 0.0], index=[10, 20, 30, 40, 50])
sf_series_index = pd.Series([1.0, 0.75, 0.5, 0.25, 0.0], index=[10, 20, 30, 40, 50])
sf_series_no_index = pd.Series([1.0, 0.75, 0.5, 0.25, 0.0])
q = 0.5
assert utils.qth_survival_times(q, sf_list) == 2
assert utils.qth_survival_times(q, sf_array) == 2
assert utils.qth_survival_times(q, sf_df_no_index) == 2
assert utils.qth_survival_times(q, sf_df_index) == 30
assert utils.qth_survival_times(q, sf_series_index) == 30
assert utils.qth_survival_times(q, sf_series_no_index) == 2
def test_qth_survival_times_with_varying_datatype_inputs():
sf_list = [1.0, 0.75, 0.5, 0.25, 0.0]
sf_array = np.array([1.0, 0.75, 0.5, 0.25, 0.0])
sf_df_no_index = pd.DataFrame([1.0, 0.75, 0.5, 0.25, 0.0])
sf_df_index = pd.DataFrame([1.0, 0.75, 0.5, 0.25, 0.0], index=[10, 20, 30, 40, 50])
sf_series_index = pd.Series([1.0, 0.75, 0.5, 0.25, 0.0], index=[10, 20, 30, 40, 50])
sf_series_no_index = pd.Series([1.0, 0.75, 0.5, 0.25, 0.0])
q = 0.5
assert utils.qth_survival_times(q, sf_list) == 2
assert utils.qth_survival_times(q, sf_array) == 2
assert utils.qth_survival_times(q, sf_df_no_index) == 2
assert utils.qth_survival_times(q, sf_df_index) == 30
assert utils.qth_survival_times(q, sf_series_index) == 30
assert utils.qth_survival_times(q, sf_series_no_index) == 2
def test_qth_survival_times_multi_dim_input():
sf = np.linspace(1, 0, 50)
sf_multi_df = pd.DataFrame({"sf": sf, "sf**2": sf ** 2})
medians = utils.qth_survival_times(0.5, sf_multi_df)
assert medians["sf"].loc[0.5] == 25
assert medians["sf**2"].loc[0.5] == 15
def qq_plot(model, ax=None, **plot_kwargs):
"""
Produces a quantile-quantile plot of the empirical CDF against
the fitted parametric CDF. Large deviances away from the line y=x
can invalidate a model (though we expect some natural deviance in the tails).
Parameters
-----------
model: obj
A fitted lifelines univariate parametric model, like ``WeibullFitter``
plot_kwargs:
kwargs for the plot.
Returns
--------
ax:
The axes which was used.
Examples
---------
>>> from lifelines import *
>>> from lifelines.plotting import qq_plot
>>> from lifelines.datasets import load_rossi
>>> df = load_rossi()
>>> wf = WeibullFitter().fit(df['week'], df['arrest'])
>>> qq_plot(wf)
"""
from lifelines.utils import qth_survival_times
from lifelines import KaplanMeierFitter
if ax is None:
ax = plt.gca()
dist = get_distribution_name_of_lifelines_model(model)
dist_object = create_scipy_stats_model_from_lifelines_model(model)
COL_EMP = "empirical quantiles"
COL_THEO = "fitted %s quantiles" % dist
if CensoringType.is_left_censoring(model):
kmf = KaplanMeierFitter().fit_left_censoring(model.durations, model.event_observed, label=COL_EMP)
elif CensoringType.is_right_censoring(model):
kmf = KaplanMeierFitter().fit_right_censoring(model.durations, model.event_observed, label=COL_EMP)
elif CensoringType.is_interval_censoring(model):
raise NotImplementedError("lifelines does not have a non-parametric interval model yet.")
q = np.unique(kmf.cumulative_density_.values[:, 0])
# this is equivalent to the old code `qth_survival_times(q, kmf.cumulative_density, cdf=True)`
quantiles = qth_survival_times(1 - q, kmf.survival_function_)
quantiles[COL_THEO] = dist_object.ppf(q)
quantiles = quantiles.replace([-np.inf, 0, np.inf], np.nan).dropna()
max_, min_ = quantiles[COL_EMP].max(), quantiles[COL_EMP].min()
quantiles.plot.scatter(COL_THEO, COL_EMP, c="none", edgecolor="k", lw=0.5, ax=ax)
ax.plot([min_, max_], [min_, max_], c="k", ls=":", lw=1.0)
ax.set_ylim(min_, max_)
ax.set_xlim(min_, max_)
return ax
def predictor(request):
house_link = request.GET['weblink']
print(request.GET['weblink'])
# get training data
feat = pd.read_csv('./survival_api_data/feat.csv')
house_all = houseScraper(house_link)
house_feat = [0]
house_feat.append(house_all['days'])
house_feat.append(house_all['discount'])
house_feat.append(house_all['price'])
house_feat.append(house_all['r2M'])
house_feat.append(house_all['MonthList'])
house_feat.append(house_all['MonthSold'])
house_feat.append(house_all['NumList'])
house_feat.append(house_all['NumPC'])
house_feat.append(house_all['NumSold'])
col_name = [
'sold', 'days', 'discount', 'price', 'r2M', 'MonthList', 'MonthSold',
'NumList', 'NumPC', 'NumSold'
]
feat_df = pd.DataFrame([house_feat], columns=col_name)
feat_all = pd.concat([feat, feat_df], ignore_index=True)
censor = house_all['days'] // 30 + 3
cph_time = get_time_model(censor, feat_all)
pred_time_0 = cph_time.predict_survival_function(feat_df)
pred_time = pred_time_0.apply(
lambda c: (c / c.loc[feat_all.loc[c.name, 'days']]).clip_upper(1))
pred_time_75 = qth_survival_times(0.25, pred_time)
pred_time_50 = qth_survival_times(0.5, pred_time)
pred_week = int((pred_time_50 - house_all['days']) // 7)
cph_off = get_off_model(5, feat_all)
pred_off_0 = cph_off.predict_survival_function(feat_df)
pred_off = pred_off_0.apply(
lambda c: (c / c.loc[feat_all.loc[c.name, 'discount']]).clip_upper(1))
pred_off_75 = qth_survival_times(0.25, pred_off)
pred_off_50 = qth_survival_times(0.5, pred_off)
if pred_off_75 != float("inf"):
off = pred_off_75
elif pred_off_50 != float("inf"):
off = pred_off_50
else:
off = 0
off_str = "{:.1f}%".format(off)
off = round(off / 100, 1)
off_usd = int(round((1 - off) * house_all['price']))
off_usd_str = '$' + format(off_usd, ',')
dom_w = int(math.ceil(house_all['days'] / 7))
dom_w_str = str(dom_w) + ' weeks'
pred_week_str = str(pred_week) + ' weeks'
response_dict = {
"log": True,
# "pred_time_75": pred_time_75,
# "pred_off_75": pred_off_75,
# "pred_off_50": pred_off_50,
"prediction": {
"address": house_all['address'],
"listing_price": house_all['listingPrice'],
"dom": dom_w_str,
"pred_weeks": pred_week_str,
"off_usd": off_usd_str,
"off_pct": off_str
}
}
response = json.dumps(response_dict)
return HttpResponse(response)
conditioned_sf = unconditioned_sf.apply(lambda c: (c/c.loc[data.loc[c.name, 'tenure']]).clip_upper(1))
# now we can investigate customers to see how the conditioning has affected their survival over the baseline rate
subject = 12
unconditioned_sf[subject].plot(ls="--", color="#A60628", label="unconditioned")
conditioned_sf[subject].plot(color="#A60628", label="conditioned on $T>58$")
plt.legend()
# we can see that cust 12 is still a customer after 58 months, which means cust 12's survival curve drops slower than
# the baseline for similar custs without that condition.
# the predict_survival_function has created a metrix of survival probabilities for each remaining customer at each
# point in time. what we need to do now is use that to select a single value as prdiction for how long a customer
# will last. Lets use the the median.
# predictions_50 = median_survival_times(conditioned_sf)
likelihood_cutoff = 0.5
predictions_50 = qth_survival_times(likelihood_cutoff, conditioned_sf) # same as above but specifying the %
# this gave us a single row with the month number where the customer has a 50% likelihood of churning.
# Lets join it to some data to investigate.
values = predictions_50.T.join(data[['MonthlyCharges', 'tenure']])
values['RemainingValue'] = values['MonthlyCharges'] * (values[likelihood_cutoff] - values['tenure'])
# now looking at the RemainingValue column we can see which customers would most affect our bottom line.
# Great, so we know which customers have the highest risk of churn and when they are likely to, but what can we do?
# lets take a look at our coefficients from earlier.
# we can see that the features that impact survival positively are 'Contract_One year', 'Contract_Two year',
# 'PaymentMethod_Bank transfer (automatic)', 'PaymentMethod_Credit card (automatic)'. Beyond these the results are
# insignificant. Lets compare customers with the features to understand the best place to spend money.
upgrades = ['Contract_One year',
'Contract_Two year',
'PaymentMethod_Bank transfer (automatic)',
def qq_plot(model, ax=None, **plot_kwargs):
"""
Produces a quantile-quantile plot of the empirical CDF against
the fitted parametric CDF. Large deviances away from the line y=x
can invalidate a model (though we expect some natural deviance in the tails).
Parameters
-----------
model: obj
A fitted lifelines univariate parametric model, like ``WeibullFitter``
plot_kwargs:
kwargs for the plot.
Returns
--------
ax:
The axes which was used.
Examples
---------
.. code:: python
from lifelines import *
from lifelines.plotting import qq_plot
from lifelines.datasets import load_rossi
df = load_rossi()
wf = WeibullFitter().fit(df['week'], df['arrest'])
qq_plot(wf)
Notes
------
The interval censoring case uses the mean between the upper and lower bounds.
"""
from lifelines.utils import qth_survival_times
from lifelines import KaplanMeierFitter
if ax is None:
ax = plt.gca()
dist = get_distribution_name_of_lifelines_model(model)
dist_object = create_scipy_stats_model_from_lifelines_model(model)
COL_EMP = "empirical quantiles"
COL_THEO = "fitted %s quantiles" % dist
if CensoringType.is_left_censoring(model):
kmf = KaplanMeierFitter().fit_left_censoring(
model.durations, model.event_observed, label=COL_EMP, weights=model.weights, entry=model.entry
)
sf, cdf = kmf.survival_function_[COL_EMP], kmf.cumulative_density_[COL_EMP]
elif CensoringType.is_right_censoring(model):
kmf = KaplanMeierFitter().fit_right_censoring(
model.durations, model.event_observed, label=COL_EMP, weights=model.weights, entry=model.entry
)
sf, cdf = kmf.survival_function_[COL_EMP], kmf.cumulative_density_[COL_EMP]
elif CensoringType.is_interval_censoring(model):
kmf = KaplanMeierFitter().fit_interval_censoring(
model.lower_bound, model.upper_bound, label=COL_EMP, weights=model.weights, entry=model.entry
)
sf, cdf = kmf.survival_function_.mean(1), kmf.cumulative_density_[COL_EMP + "_lower"]
q = np.unique(cdf.values)
quantiles = qth_survival_times(1 - q, sf)
quantiles[COL_THEO] = dist_object.ppf(q)
quantiles = quantiles.replace([-np.inf, 0, np.inf], np.nan).dropna()
max_, min_ = quantiles[COL_EMP].max(), quantiles[COL_EMP].min()
quantiles.plot.scatter(COL_THEO, COL_EMP, c="none", edgecolor="k", lw=0.5, ax=ax)
ax.plot([min_, max_], [min_, max_], c="k", ls=":", lw=1.0)
ax.set_ylim(min_, max_)
ax.set_xlim(min_, max_)
return ax
def plot_survival(unique_groups, grouped_data, analysis_type, censors, ci, showplot, stat_results, time='Months'):
#plot survival curve
kmf = KaplanMeierFitter()
fig, ax = plt.subplots()
n_in_groups = []
f = open('Kaplan_%s.txt' % (analysis_type), 'a')
f.write("\nPercent %s\n" % analysis_type)
headers = "Group\t"
for x in range(95,-1,-5):
headers += str(x) + "%\t"
f.write("%s\n" % headers)
for i, group in enumerate(unique_groups):
data = grouped_data.get_group(group)
n_in_groups.append(len(data))
# Adjust survival data from days to whatever form wanted
if time.lower() == 'months':
survival_time = (data['survival']/(365/12))
elif time.lower() == 'years':
survival_time = (data['survival']/(365))
else:
survival_time = data['survival']
kmf.fit(survival_time, data['event'], label = group)
# print(data[survival])
# print(kmf.survival_function_)
f.write("%s\t" % group)
for x in range(95, -1, -5):
f.write(str(qth_survival_times(x/100, kmf.survival_function_)) + "\t")
f.write("\n")
kmf.plot(ax=ax, show_censors=censors, ci_show=ci, linewidth=2.5)
# Make the graph pretty!
textbox = dict(horizontalalignment = 'left', verticalalignment = 'bottom', fontname = 'Arial', fontsize = 18)
labels = dict(horizontalalignment = 'center', verticalalignment = 'center', fontname = 'Arial', fontsize = 28)
ax.grid(False)
ax.set_ylim(0,1.05)
ax.spines['left'].set_linewidth(2.5)
ax.spines['right'].set_linewidth(2.5)
ax.spines['top'].set_linewidth(2.5)
ax.spines['bottom'].set_linewidth(2.5)
ax.yaxis.set_tick_params(width=2.5)
ax.xaxis.set_tick_params(width=2.5)
ax.xaxis.set_ticks_position('bottom')
ax.yaxis.set_ticks_position('left')
# plt.title('%s' % (analysis_type), labels, y = 1.05)
plt.xlabel('%s Post-Diagnosis' % time, labels, labelpad = 20)
if analysis_type == 'survival':
plt.ylabel('Overall Survival', labels, labelpad = 20)
else:
plt.ylabel('Relapse-Free Survival', labels, labelpad=20)
plt.xticks(fontname = 'Arial', fontsize = 24)
plt.yticks(fontname = 'Arial', fontsize = 24)
ax.tick_params(axis='y', pad=10)
ax.tick_params(axis='x', pad=10)
legend = ax.legend(frameon=False,loc=3)
counter=0
for label in legend.get_texts():
label.set_fontsize(20)
label.set_text('%s n=%d' % (unique_groups[counter], n_in_groups[counter]))
counter += 1
if len(unique_groups) == 2:
plt.text(0.95, 0.05, 'p = %.2g' % (stat_results.p_value), fontname='Arial', fontsize=20, ha='right', transform=ax.transAxes)
plt.tight_layout()
fig.savefig('Kaplan_%s.png' % analysis_type, transparent = True)
fig.savefig('Kaplan_%s.eps' % analysis_type, transparent = True)
if showplot == True:
plt.show()
plt.close(fig)
def test_qth_survival_times_multi_dim_input():
sf = np.linspace(1, 0, 50)
sf_multi_df = pd.DataFrame({'sf': sf, 'sf**2': sf ** 2})
medians = utils.qth_survival_times(0.5, sf_multi_df)
assert medians.ix['sf'][0.5] == 25
assert medians.ix['sf**2'][0.5] == 15
def test_qth_survival_times_multi_dim_input():
sf = np.linspace(1, 0, 50)
sf_multi_df = pd.DataFrame({'sf': sf, 'sf**2': sf**2})
medians = utils.qth_survival_times(0.5, sf_multi_df)
assert medians.ix['sf'][0.5] == 25
assert medians.ix['sf**2'][0.5] == 15
def predict_percentile(self, df, p=0.5):
return qth_survival_times(p, self.predict_survival_function(df))
def test_qth_survival_time_with_cdf_instead_of_survival_function():
cdf = np.linspace(0, 1, 50)
assert utils.qth_survival_times(0.5, cdf, cdf=True) == 25
def qq_plot(model, **plot_kwargs):
"""
Produces a quantile-quantile plot of the empirical CDF against
the fitted parametric CDF. Large deviances away from the line y=x
can invalidate a model (though we expect some natural deviance in the tails).
Parameters
-----------
model: obj
A fitted lifelines univariate parametric model, like ``WeibullFitter``
plot_kwargs:
kwargs for the plot.
Returns
--------
ax: axis object
Examples
---------
>>> from lifelines import *
>>> from lifelines.plotting import qq_plot
>>> from lifelines.datasets import load_rossi
>>> df = load_rossi()
>>> wf = WeibullFitter().fit(df['week'], df['arrest'])
>>> qq_plot(wf)
"""
from lifelines.utils import qth_survival_times
from lifelines import KaplanMeierFitter
from lifelines.fitters import KnownModelParametericUnivariateFitter
assert isinstance(model, KnownModelParametericUnivariateFitter)
set_kwargs_ax(plot_kwargs)
ax = plot_kwargs.pop("ax")
dist = get_distribution_name_of_lifelines_model(model)
dist_object = create_scipy_stats_model_from_lifelines_model(model)
COL_EMP = "empirical quantiles"
COL_THEO = "fitted %s quantiles" % dist
kmf = KaplanMeierFitter().fit(model.durations,
model.event_observed,
left_censorship=model.left_censorship,
label=COL_EMP)
if model.left_censorship:
q = np.unique(kmf.cumulative_density_.values[:, 0])
quantiles = qth_survival_times(q, kmf.cumulative_density_, cdf=True)
else:
q = np.unique(1 - kmf.survival_function_.values[:, 0])
quantiles = qth_survival_times(q, 1 - kmf.survival_function_, cdf=True)
quantiles[COL_THEO] = dist_object.ppf(q)
quantiles = quantiles.replace([-np.inf, 0, np.inf], np.nan).dropna()
max_, min_ = quantiles[COL_EMP].max(), quantiles[COL_EMP].min()
quantiles.plot.scatter(COL_THEO,
COL_EMP,
c="none",
edgecolor="k",
lw=0.5,
ax=ax)
ax.plot([min_, max_], [min_, max_], c="k", ls=":", lw=1.0)
ax.set_ylim(min_, max_)
ax.set_xlim(min_, max_)
return ax
[1000, 10000, 0, 0, 1, 0, 2], [1000, 10000, 0, 1, 1, 0, 2]]
num_d = len(d)
dfn = pd.DataFrame(
d, columns=["ID", "KM", "DEAD", "ENGINE", "MOUNTAIN", "CITY", "MONDAY"])
print(dfn)
censored_subjects = censored_subjects.append(dfn, ignore_index=True)
print(censored_subjects)
unconditioned_sf = cph.predict_survival_function(censored_subjects)
print(unconditioned_sf)
from lifelines.utils import median_survival_times, qth_survival_times
predictions_75 = qth_survival_times(0.75, unconditioned_sf)
predictions_25 = qth_survival_times(0.25, unconditioned_sf)
predictions_50 = median_survival_times(unconditioned_sf)
print(predictions_50)
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(8, 4))
for f in unconditioned_sf:
ax.plot(unconditioned_sf[f], alpha=.5, label=f)
#ax.legend()
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(8, 4))
for i, f in enumerate(reversed(unconditioned_sf.columns)):
#print( i )
if i < num_d:
print(i, f)
ax.plot(unconditioned_sf[f], alpha=1, label=f)