def comb_cdf(x1, x2, **kwargs):
""" Compute the combined CDF """
""" x12 = x1 U x2, f1 = F1(x12), f2 = F2(x12) """
""" Input """
""" x1: Array of double """
""" x2: Array of double """
""" **kwargs: issorted - True or False. Default is False """
""" Return: (x12, f1, f2) """
isSortedVal = False
mtype = "dice"
n1 = float(len(x1))
n2 = float(len(x2))
if "issorted" in kwargs:
isSortedVal = kwargs["issorted"]
if "mtype" in kwargs:
mtype = kwargs["mtype"]
if isSortedVal == False:
x1.sort()
x2.sort()
x12max = min(x1.max(), x2.max())
sx12 = set(x1).union(set(x2))
px12 = np.array(list(sx12))
px12.sort()
x12 = remove_outliers(px12, 0.0, x12max)
n12 = len(x12)
if False:
f1 = np.zeros(n12)
f2 = np.zeros(n12)
i = 0
while i < n12:
xij = x12[i]
idx1 = np.argwhere(x1 <= xij)
if (len(idx1) > 0):
f1[i] = (idx1.max() + 1) / (n1 + 1.0)
else:
f1[i] = 0.0
idx2 = np.argwhere(x2 <= xij)
if (len(idx2) > 0):
f2[i] = (idx2.max() + 1) / (n2 + 1.0)
else:
f2[i] = 0.0
i += 1
else:
pass
f1 = np.array(
[len(np.argwhere(x1 <= xij)) / float(n1 + 1.0) for xij in x12])
f2 = np.array(
[len(np.argwhere(x2 <= xij)) / float(n2 + 1.0) for xij in x12])
return (x12, f1, f2)
def comb_cdf(x1, x2, **kwargs): """ Compute the combined CDF """ """ x12 = x1 U x2, f1 = F1(x12), f2 = F2(x12) """ """ Input """ """ x1: Array of double """ """ x2: Array of double """ """ **kwargs: issorted - True or False. Default is False """ """ Return: (x12, f1, f2) """ isSortedVal = False mtype = "dice" n1 = float(len(x1)) n2 = float(len(x2)) if "issorted" in kwargs: isSortedVal = kwargs["issorted"] if "mtype" in kwargs: mtype = kwargs["mtype"] if isSortedVal == False: x1.sort() x2.sort() x12max = min(x1.max(), x2.max()) sx12 = set(x1).union(set(x2)) px12 = np.array(list(sx12)) px12.sort() x12 = remove_outliers(px12,0.0, x12max) n12 = len(x12) if False: f1 = np.zeros(n12) f2 = np.zeros(n12) i = 0 while i < n12: xij = x12[i] idx1 = np.argwhere(x1 <= xij) if ( len(idx1) > 0 ): f1[i] = (idx1.max()+1)/(n1 + 1.0) else: f1[i] = 0.0 idx2 = np.argwhere(x2 <= xij) if ( len(idx2) > 0 ): f2[i] = (idx2.max()+1)/(n2 + 1.0) else: f2[i] = 0.0 i+= 1 else: pass f1 = np.array([len(np.argwhere(x1 <= xij))/float(n1 + 1.0) for xij in x12]) f2 = np.array([len(np.argwhere(x2 <= xij))/float(n2 + 1.0) for xij in x12]) return (x12,f1,f2)
def simple_dice(xs, ys):
r = SystemRandom()
x = remove_outliers(np.array(xs), 0, np.max(xs))
y = remove_outliers(np.array(ys), 0, np.max(ys))
x.sort()
y.sort()
nx = len(x)
ny = len(y)
darray = np.zeros(50)
for i in xrange(0, 50):
if nx < ny:
xprime = x
yprime = r.sample(y, nx)
n = float(nx)
else:
xprime = r.sample(x, ny)
yprime = y
n = float(ny)
xprime.sort()
yprime.sort()
darray[i] = kumar_hassebrook(xprime, yprime)
mu = np.mean(darray)
sigma = np.std(darray)
conf_level = 0.99
alpha = 1 - conf_level
alpha = 1.0 - conf_level
z_alpha_2 = norm.ppf(1 - (alpha / 2.0))
e = z_alpha_2 * sigma / math.sqrt(len(darray))
return mu + e
def simple_dice(xs, ys): r = SystemRandom() x = remove_outliers(np.array(xs), 0, np.max(xs)) y = remove_outliers(np.array(ys), 0, np.max(ys)) x.sort() y.sort() nx = len(x) ny = len(y) darray = np.zeros(50) for i in xrange(0,50): if nx < ny: xprime = x yprime = r.sample(y, nx) n = float(nx) else: xprime = r.sample(x, ny) yprime = y n = float(ny) xprime.sort() yprime.sort() darray[i] = kumar_hassebrook(xprime,yprime) mu = np.mean(darray) sigma = np.std(darray) conf_level = 0.99 alpha = 1-conf_level alpha = 1.0 - conf_level z_alpha_2 = norm.ppf(1-(alpha/2.0)) e = z_alpha_2 * sigma/math.sqrt(len(darray)) return mu+e
