scipy-yli/yli/survival.py

#   scipy-yli: Helpful SciPy utilities and recipes
#   Copyright © 2022–2023  Lee Yingtong Li (RunasSudo)
#
#   This program is free software: you can redistribute it and/or modify
#   it under the terms of the GNU Affero General Public License as published by
#   the Free Software Foundation, either version 3 of the License, or
#   (at your option) any later version.
#
#   This program is distributed in the hope that it will be useful,
#   but WITHOUT ANY WARRANTY; without even the implied warranty of
#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#   GNU Affero General Public License for more details.
#
#   You should have received a copy of the GNU Affero General Public License
#   along with this program.  If not, see <https://www.gnu.org/licenses/>.

import numpy as np
import pandas as pd
from scipy import stats
import statsmodels.api as sm

import io
import json
import subprocess

from .config import config
from .sig_tests import ChiSquaredResult
from .utils import Estimate, check_nan

def kaplanmeier(df, time, status, by=None, *, ci=True, transform_x=None, transform_y=None, nan_policy='warn', fig=None, ax=None):
	"""
	Generate a Kaplan–Meier plot
	
	Uses the Python *matplotlib* library.
	
	:param df: Data to generate plot for
	:type df: DataFrame
	:param time: Column in *df* for the time to event (numeric or timedelta)
	:type time: str
	:param status: Column in *df* for the status variable (True/False or 1/0)
	:type status: str
	:param by: Column in *df* to stratify by (categorical)
	:type by: str
	:param ci: Whether to plot confidence intervals around the survival function
	:type ci: bool
	:param transform_x: Function to transform x axis by
	:type transform_x: callable
	:param transform_y: Function to transform y axis by
	:type transform_y: callable
	:param nan_policy: How to handle *nan* values (see :ref:`nan-handling`)
	:type nan_policy: str
	
	:rtype: (Figure, Axes)
	"""
	
	import matplotlib.pyplot as plt
	
	# Check for/clean NaNs
	if by:
		df = check_nan(df[[time, status, by]], nan_policy)
	else:
		df = check_nan(df[[time, status]], nan_policy)
	
	# Covert timedelta to numeric
	df, time_units = survtime_to_numeric(df, time)
	
	if ax is None:
		fig, ax = plt.subplots()
	
	if by is not None:
		# Group by independent variable
		groups = df.groupby(by)
		
		for group in groups.groups:
			subset = groups.get_group(group)
			handle = plot_survfunc_kaplanmeier(ax, subset[time], subset[status], ci, transform_x, transform_y)
			handle.set_label('{} = {}'.format(by, group))
	else:
		# No grouping
		plot_survfunc_kaplanmeier(ax, df[time], df[status], ci, transform_x, transform_y)
	
	if time_units:
		ax.set_xlabel('{} ({})'.format(time, time_units))
	else:
		ax.set_xlabel(time)
	ax.set_ylabel('Survival probability ({:.0%} CI)'.format(1-config.alpha) if ci else 'Survival probability')
	ax.set_xlim(left=0)
	ax.set_ylim(0, 1)
	
	if by is not None:
		ax.legend()
	
	return fig, ax

def plot_survfunc_kaplanmeier(ax, time, status, ci, transform_x=None, transform_y=None):
	xpoints, ypoints, ypoints0, ypoints1 = calc_survfunc_kaplanmeier(time, status, ci, transform_x, transform_y)
	
	handle = ax.plot(xpoints, ypoints)[0]
	
	if ci:
		ax.fill_between(xpoints, ypoints0, ypoints1, alpha=0.3, label='_')
	
	return handle

def calc_survfunc_kaplanmeier(time, status, ci, transform_x=None, transform_y=None):
	# Estimate the survival function
	sf = sm.SurvfuncRight(time, status)
	
	# Draw straight lines
	# np.concatenate(...) to force starting drawing from time 0, survival 100%
	xpoints = np.concatenate([[0], sf.surv_times]).repeat(2)[1:]
	ypoints = np.concatenate([[1], sf.surv_prob]).repeat(2)[:-1]
	
	if transform_x:
		xpoints = transform_x(xpoints)
	if transform_y:
		ypoints = transform_y(ypoints)
	
	if ci:
		zstar = -stats.norm.ppf(config.alpha/2)
		
		# Get confidence intervals
		ci0 = sf.surv_prob - zstar * sf.surv_prob_se
		ci1 = sf.surv_prob + zstar * sf.surv_prob_se
		
		# Plot confidence intervals
		ypoints0 = np.concatenate([[1], ci0]).repeat(2)[:-1]
		ypoints1 = np.concatenate([[1], ci1]).repeat(2)[:-1]
		
		if transform_y:
			ypoints0 = transform_y(ypoints0)
			ypoints1 = transform_y(ypoints1)
		
		return xpoints, ypoints, ypoints0, ypoints1
	
	return xpoints, ypoints, None, None

def turnbull(df, time_left, time_right, by=None, *, ci=True, step_loc=0.5, maxiter=None, ll_tolerance=None, se_method=None, zero_tolerance=None, ci_precision=None, transform_x=None, transform_y=None, nan_policy='warn', fig=None, ax=None):
	"""
	Generate a Turnbull estimator plot, which extends the Kaplan–Meier estimator to interval-censored observations
	
	The intervals are assumed to be half-open intervals, (*left*, *right*]. *right* == *np.inf* implies the event was right-censored.
	
	By default, the survival function is drawn as a step function at the midpoint of each Turnbull interval.
	
	Uses the hpstat *turnbull* command.
	
	:param df: Data to generate plot for
	:type df: DataFrame
	:param time_left: Column in *df* for the time to event, left interval endpoint (numeric or timedelta)
	:type time_left: str
	:param time_right: Column in *df* for the time to event, right interval endpoint (numeric or timedelta)
	:type time_right: str
	:param by: Column in *df* to stratify by (categorical)
	:type by: str
	:param ci: Whether to plot confidence intervals around the survival function
	:type ci: bool
	:param step_loc: Proportion along the length of each Turnbull interval to step down the survival function, e.g. 0 for left bound, 1 for right bound, 0.5 for interval midpoint
	:type step_loc: float
	:param maxiter: Maximum number of iterations to attempt
	:type maxiter: int
	:param ll_tolerance: Terminate algorithm when the absolute change in log-likelihood is less than this tolerance
	:type ll_tolerance: float
	:param se_method: Method for computing standard error or survival probabilities (see hpstat *turnbull* documentation)
	:type se_method: str
	:param zero_tolerance: Threshold for dropping failure probability when se_method is "oim-drop-zeros"
	:type zero_tolerance: float
	:param ci_precision: Desired precision of confidence limits when se-method is "likelihood-ratio"
	:type ci_precision: float
	:param transform_x: Function to transform x axis by
	:type transform_x: callable
	:param transform_y: Function to transform y axis by
	:type transform_y: callable
	:param nan_policy: How to handle *nan* values (see :ref:`nan-handling`)
	:type nan_policy: str
	
	:rtype: (Figure, Axes)
	"""
	
	import matplotlib.pyplot as plt
	
	# Check for/clean NaNs
	if by:
		df = check_nan(df[[time_left, time_right, by]], nan_policy)
	else:
		df = check_nan(df[[time_left, time_right]], nan_policy)
	
	# Covert timedelta to numeric
	df, time_units = survtime_to_numeric(df, time_left, time_right)
	
	if ax is None:
		fig, ax = plt.subplots()
	
	if by is not None:
		# Group by independent variable
		groups = df.groupby(by)
		
		for group in groups.groups:
			subset = groups.get_group(group)
			handle = plot_survfunc_turnbull(
				ax, subset[time_left], subset[time_right],
				ci=ci, step_loc=step_loc, maxiter=maxiter, ll_tolerance=ll_tolerance, se_method=se_method, zero_tolerance=zero_tolerance, ci_precision=ci_precision, transform_x=transform_x, transform_y=transform_y
			)
			handle.set_label('{} = {}'.format(by, group))
	else:
		# No grouping
		plot_survfunc_turnbull(
			ax, df[time_left], df[time_right],
			ci=ci, step_loc=step_loc, maxiter=maxiter, ll_tolerance=ll_tolerance, se_method=se_method, zero_tolerance=zero_tolerance, ci_precision=ci_precision, transform_x=transform_x, transform_y=transform_y
		)
	
	if time_units:
		ax.set_xlabel('{} + {} ({})'.format(time_left, time_right, time_units))
	else:
		ax.set_xlabel('{} + {}'.format(time_left, time_right))
	ax.set_ylabel('Survival probability')
	ax.set_xlim(left=0)
	ax.set_ylim(0, 1)
	
	if by is not None:
		ax.legend()
	
	return fig, ax

def plot_survfunc_turnbull(ax, time_left, time_right, *, ci=True, step_loc=0.5, maxiter=None, ll_tolerance=None, se_method=None, zero_tolerance=None, ci_precision=None, transform_x=None, transform_y=None):
	xpoints, ypoints, ypoints0, ypoints1 = calc_survfunc_turnbull(
		time_left, time_right,
		ci=ci, step_loc=step_loc, maxiter=maxiter, ll_tolerance=ll_tolerance, se_method=se_method, zero_tolerance=zero_tolerance, ci_precision=ci_precision, transform_x=transform_x, transform_y=transform_y
	)
	
	handle = ax.plot(xpoints, ypoints)[0]
	
	if ci:
		ax.fill_between(xpoints, ypoints0, ypoints1, alpha=0.3, label='_')
	
	return handle

def calc_survfunc_turnbull(time_left, time_right, *, ci=True, step_loc=0.5, maxiter=None, ll_tolerance=None, se_method=None, zero_tolerance=None, ci_precision=None, transform_x=None, transform_y=None):
	# Estimate the survival function
	
	# Prepare arguments
	hpstat_args = [config.hpstat_path, 'turnbull', '-', '--output', 'json']
	if maxiter:
		hpstat_args.append('--max-iterations')
		hpstat_args.append(str(maxiter))
	if ll_tolerance:
		hpstat_args.append('--ll-tolerance')
		hpstat_args.append(str(ll_tolerance))
	if se_method:
		hpstat_args.append('--se-method')
		hpstat_args.append(se_method)
	elif not ci:
		hpstat_args.append('--se-method')
		hpstat_args.append('none')
	if zero_tolerance:
		hpstat_args.append('--zero-tolerance')
		hpstat_args.append(str(zero_tolerance))
	if ci_precision:
		hpstat_args.append('--ci-precision')
		hpstat_args.append(str(ci_precision))
	
	# Export data to CSV
	csv_buf = io.StringIO()
	pd.DataFrame({'LeftTime': time_left, 'RightTime': time_right}).to_csv(csv_buf, index=False)
	csv_str = csv_buf.getvalue()
	
	# Run hpstat binary
	proc = subprocess.run(hpstat_args, input=csv_str, stdout=subprocess.PIPE, stderr=None, encoding='utf-8', check=True)
	raw_result = json.loads(proc.stdout)
	
	survival_prob = np.array(raw_result['survival_prob'])
	
	from IPython.display import clear_output
	clear_output(wait=True)
	
	xpoints = [i[0]*(1-step_loc) + i[1]*step_loc for i in raw_result['failure_intervals'] if i[1]]
	ypoints = survival_prob
	if raw_result['failure_intervals'][-1][1]:
		# No right-censored observations - we can draw the whole survival curve
		ypoints = np.concatenate([ypoints, [0]])
	
	# Draw straight lines
	# np.concatenate(...) to force starting drawing from time 0, survival 100%
	xpoints = np.concatenate([[0], xpoints]).repeat(2)[1:]
	ypoints = np.concatenate([[1], ypoints]).repeat(2)[:-1]
	
	if transform_x:
		xpoints = transform_x(xpoints)
	if transform_y:
		ypoints = transform_y(ypoints)
	
	if ci:
		# Get confidence intervals
		if raw_result['survival_prob_se']:
			zstar = -stats.norm.ppf(config.alpha/2)
			survival_prob_se = np.array(raw_result['survival_prob_se'])
			
			ci0 = survival_prob - zstar * survival_prob_se
			ci1 = survival_prob + zstar * survival_prob_se
		else:
			survival_prob_ci = np.array(raw_result['survival_prob_ci'])
			ci0 = survival_prob_ci.T[0]
			ci1 = survival_prob_ci.T[1]
		
		if raw_result['failure_intervals'][-1][1]:
			# No right-censored observations - we can draw the whole survival curve
			ci0 = np.concatenate([ci0, [0]])
			ci1 = np.concatenate([ci1, [0]])
		
		# Plot confidence intervals
		ypoints0 = np.concatenate([[1], ci0]).repeat(2)[:-1]
		ypoints1 = np.concatenate([[1], ci1]).repeat(2)[:-1]
		
		if transform_y:
			ypoints0 = transform_y(ypoints0)
			ypoints1 = transform_y(ypoints1)
		
		return xpoints, ypoints, ypoints0, ypoints1
	
	return xpoints, ypoints, None, None

def survtime_to_numeric(df, time, time2=None):
	"""
	Convert pandas timedelta dtype to float64, auto-detecting the best time unit to display
	
	:param df: Data to check for pandas timedelta dtype
	:type df: DataFrame
	:param time: Column to check for pandas timedelta dtype
	:type df: DataFrame
	:param time: Second column, if any, to check for pandas timedelta dtype
	:type df: DataFrame
	
	:return: (*df*, *time_units*)
		
		* **df** (*DataFrame*) – Data with pandas timedelta dtypes converted, which is *not* copied
		* **time_units** (*str*) – Human-readable description of the time unit, or *None* if not converted
	"""
	
	max_time = None
	
	if df[time].dtype == '<m8[ns]':
		df[time] = df[time].dt.total_seconds()
		max_time = df[time].max()
	
	if time2 and df[time2].dtype == '<m8[ns]':
		df[time2] = df[time2].dt.total_seconds()
		max_time = max(max_time or 0, df[time2].max())
	
	if max_time is not None:
		# Auto-detect best time units
		if max_time > 365.24*24*60*60:
			time_divider = 365.24*24*60*60
			time_units = 'years'
		elif max_time > 7*24*60*60 / 12:
			time_divider = 7*24*60*60
			time_units = 'weeks'
		elif max_time > 24*60*60:
			time_divider = 24*60*60
			time_units = 'days'
		elif max_time > 60*60:
			time_divider = 60*60
			time_units = 'hours'
		elif max_time > 60:
			time_divider = 60
			time_units = 'minutes'
		else:
			time_divider = 1
			time_units = 'seconds'
		
		df[time] /= time_divider
		
		if time2:
			df[time2] /= time_divider
		
		return df, time_units
	else:
		return df, None

def logrank(df, time, status, by, nan_policy='warn'):
	"""
	Perform the log-rank test for equality of survival functions
	
	:param df: Data to perform the test on
	:type df: DataFrame
	:param time: Column in *df* for the time to event (numeric or timedelta)
	:type time: str
	:param status: Column in *df* for the status variable (True/False or 1/0)
	:type status: str
	:param by: Column in *df* to stratify by (categorical)
	:type by: str
	:param nan_policy: How to handle *nan* values (see :ref:`nan-handling`)
	:type nan_policy: str
	
	:rtype: :class:`yli.sig_tests.ChiSquaredResult`
	"""
	
	# TODO: Example
	
	# Check for/clean NaNs
	df = check_nan(df[[time, status, by]], nan_policy)
	
	if df[time].dtype == '<m8[ns]':
		df[time] = df[time].dt.total_seconds()
	
	statistic, pvalue = sm.duration.survdiff(df[time], df[status], df[by])
	
	return ChiSquaredResult(statistic=statistic, dof=1, pvalue=pvalue)