def bootstrap_ttest_multi(bfg, tails='two'):
st = []
for d1 in bfg:
for d2 in bfg:
diff = bfg[d1] - bfg[d2]
pct = scipy.stats.percentileofscore(diff, 0, kind='mean') / 100.
p = min(pct, 1 - pct)
if tails == 'two': p *= 2
star = stats.get_star(p)
st.append([d1, d2, np.mean(diff), p, star])
st = pandas.DataFrame(st, columns=['var1', 'var2', 'mean', 'p', 'sig'])
return st
def bootstrap_ttest_multi(bfg, tails='two'):
st = []
for d1 in bfg:
for d2 in bfg:
diff = bfg[d1] - bfg[d2]
pct = scipy.stats.percentileofscore(diff, 0, kind='mean') / 100.
p = min(pct, 1-pct)
if tails == 'two': p *= 2
star = stats.get_star(p)
st.append([d1, d2, np.mean(diff), p, star])
st = pandas.DataFrame(st, columns=['var1', 'var2', 'mean', 'p', 'sig'])
return st
def bootstrap_ttest_grouped(self, bf, tails="one"):
bfg = bf.groupby(["depth", "iter"]).mean()
bfg = bfg.unstack(level="depth").bootstrap
st = []
for nd1, d1 in enumerate(bfg):
for d2 in bfg.iloc[:, nd1 + 1 :]:
diff = np.squeeze(bfg[d1].values - bfg[d2].values)
pct = scipy.stats.percentileofscore(diff, 0, kind="mean") / 100.0
p = min(pct, 1 - pct)
if tails == "two":
p *= 2
star = stats.get_star(p)
st.append([d1, d2, np.mean(diff), p, star])
st = pandas.DataFrame(st, columns=["depth1", "depth2", "mean", "p", "sig"])
print(st)
return st
def bootstrap_ttest_grouped(self, bf, tails='one'):
bfg = bf.groupby(['depth', 'iter']).mean()
bfg = bfg.unstack(level='depth').bootstrap
st = []
for nd1, d1 in enumerate(bfg):
for d2 in bfg.iloc[:, nd1 + 1:]:
diff = np.squeeze(bfg[d1].values - bfg[d2].values)
pct = scipy.stats.percentileofscore(diff, 0,
kind='mean') / 100.
p = min(pct, 1 - pct)
if tails == 'two': p *= 2
star = stats.get_star(p)
st.append([d1, d2, np.mean(diff), p, star])
st = pandas.DataFrame(st,
columns=['depth1', 'depth2', 'mean', 'p', 'sig'])
print(st)
return st
