def bootstrap_mean_formatter(series, q=0.95):
if q < 0 or q > 1:
raise ValueError("Invalid q %0.2f, needs to be within [0, 1]" % q)
q = q * 100
value = series.mean()
low = np.percentile(series, (100 - q) / 2.0)
high = np.percentile(series, (q + ((100 - q) / 2.0)))
return ("%s, %d%% CI (%s, %s)" % (float_str(value), int(q), float_str(low), float_str(high)))
def hr_posterior_formatter(series, q=0.95, n=None,
summary='mean',
q_format="{q}% CI ({low}, {high})",
stat='HR',
stat_format="{stat}={value}",
include_p=False, p_compare=0):
# construct resulting string
string_parts = list()
if n:
string_parts.append("n=%d" % (int(n)))
# summarize stat if summary is given as a list
if isinstance(summary, list):
for summary_type in summary:
if summary_type == 'mean':
value = series.mean()
elif summary_type == 'median':
value = series.median()
else:
raise ValueError("invalid summary %s; needs to be either None, 'mean', or 'median'" % summary_type)
string_parts.append(stat_format.format(stat='{} ({})'.format(stat, summary_type),
value=float_str(value))
)
# summarize stat if summary is given as a single str
elif isinstance(summary, str):
if summary == 'mean':
value = series.mean()
elif summary == 'median':
value = series.median()
else:
raise ValueError("invalid summary %s; needs to be either None, 'mean', or 'median'" % summary)
string_parts.append(stat_format.format(stat=stat, value=float_str(value)))
# add posterior interval if q is given
if q:
if q < 0 or q > 1:
raise ValueError("Invalid q %0.2f, needs to be within [0, 1]" % q)
q = q * 100
low = np.percentile(series, (100 - q) / 2.0)
high = np.percentile(series, (q + ((100 - q) / 2.0)))
string_parts.append(q_format.format(q=int(q), low=float_str(low), high=float_str(high)))
# compute 'bayesian p-value', ie probability that HR <> 0
if include_p:
prob = 1-np.mean(series >= p_compare)
direction = '>'
if prob >= 0.5:
prob = 1-prob
direction = '<'
string_parts.append('p({}{}{})={}'.format(stat, direction, p_compare, float_str(prob)))
if len(string_parts) == 0:
raise ValueError('No summary components specified')
return ', '.join(string_parts)
def test_rounding():
for f, expected_str in ROUNDING_CASES.items():
eq_(expected_str, float_str(f))
def test_rounding():
for f, expected_str in ROUNDING_CASES.items():
eq_(expected_str, float_str(f))
def round_func(val):
if round_to_int:
return format(int(round(val)), ",d")
else:
return float_str(val)
def spearmanr_formatter(results):
return "n=%d, Spearman rho=%s p=%s" % (len(results.series_x), float_str(results.coeff), float_str(results.p_value))
def pearsonr_formatter(results):
return "n=%d, Pearson r=%s p=%s" % (len(results.series_x), float_str(results.coeff), float_str(results.p_value))
def logrank_formatter(results):
return "n=%d, log-rank p=%s" % (
len(results.with_condition_series) + len(results.without_condition_series), float_str(results.p_value))
def fishers_exact_formatter(results):
return "n=%d, Fisher's Exact p=%s" % (len(results.with_condition1_series) + len(results.without_condition1_series), float_str(results.p_value))
def mann_whitney_formatter(results):
return "n=%d, Mann-Whitney p=%s" % (len(results.with_condition_series) + len(results.without_condition_series), float_str(results.p_value))