Delete Bootstrapping.py

Bootstrapping.py

-from HelperFunctions import FakeFuncCode, SelectRangeNumba, EmpiricalPPF
-from KDEpy import FFTKDE
-import numpy as np
-from Model import epsilon
-from scipy.interpolate import interp1d
-from iminuit import Minuit
-from astropy.stats import bootstrap, scott_bin_width
-import scipy.special as sc
-import logging
-
-            
-    
-def Bootstrap(data, statistic, n_bootstrap, alpha=0.95, ):
-    if not (0 < alpha < 1):
-        raise ValueError("confidence level must be in (0, 1)")
-
-
-    if len(data) < 1:
-        raise ValueError("data must contain at least one measurement.")
-
-
-    boot_stat = bootstrap(data, n_bootstrap, bootfunc=statistic)
-
-    stat_data = statistic(data)
-    mean_stat = np.mean(boot_stat)
-    est_stat = 2*stat_data - mean_stat
-    std_err = np.std(boot_stat)
-    z_score = np.sqrt(2.0)*sc.erfinv(alpha)
-    conf_interval = est_stat + z_score*np.array((-std_err, std_err))
-
-
-    return est_stat, std_err, conf_interval
-    
-    
-    
-    
-def BootstrapKDE(data,
-                 n_bootstrap,
-                 n_call=1000,
-                 n_iterations=10,
-                 kernel = "gaussian",
-                 n_kde_samples=2**14,
-                 alpha=0.95,
-                 bw_limits=(epsilon(), None),
-                 limits=None,
-                 verbose = False,
-                 bw = "ISJ"
-
-                ):
-
-    if not (0 < alpha < 1):
-        raise ValueError("Bootstrap confidence level, alpha, must be in (0, 1).")
-
-    if len(data) <= 3:
-        raise ValueError("Bootstrap data must contain at least three measurements.")
-        
-        print(verbose)
-
-
-
-    kde = FFTKDE(kernel = kernel, bw=bw).fit(data)
-
-    x_kde, y_kde_orig = kde.evaluate(n_kde_samples)
-
-
-    boot_data = bootstrap(data, n_bootstrap)
-    y_kde_bs = np.vstack([FFTKDE(kernel = kernel, bw=bw).fit(_data).evaluate(x_kde)
-                               for _data in boot_data])
-
-
-    y_kde_mean = np.mean(y_kde_bs, axis=0)
-    y_kde = 2*y_kde_orig - y_kde_mean
-    y_kde_err = np.std(y_kde_bs, axis=0)
-    z_score = np.sqrt(2.0)*sc.erfinv(alpha)
-    y_kde_conf_interval = y_kde + z_score*np.array((-y_kde_err, y_kde_err))
-
-    if(limits is not None):
-        cond_inrange =(x_kde>np.min(limits)) & (x_kde<np.max(limits))
-        x_kde = x_kde[cond_inrange]
-        y_kde = y_kde[cond_inrange]
-        y_kde_err = y_kde_err[cond_inrange]
-
-    return x_kde, y_kde, y_kde_err, y_kde_conf_interval
-
-
-def Bootstrap(data, statistic, n_bootstrap, alpha=0.95):
-    if not (0 < alpha < 1):
-        raise ValueError("Confidence level must be in (0, 1)")
-
-
-    if len(data) < 1:
-        raise ValueError("Data must contain at least one measurement.")
-
-
-    boot_stat = bootstrap(data, n_bootstrap, bootfunc=statistic)
-
-    stat_data = statistic(data)
-    mean_stat = np.mean(boot_stat)
-    est_stat = 2*stat_data - mean_stat
-    std_err = np.std(boot_stat)
-    z_score = np.sqrt(2.0)*sc.erfinv(alpha)
-    conf_interval = est_stat + z_score*np.array((-std_err, std_err))
-
-
-    return est_stat, std_err, conf_interval