def _power_analysis(self):
n_required = sms.NormalIndPower().solve_power(self.effect_size,
power=self.beta,
alpha=self.alpha,
ratio=1)
n_required = ceil(n_required)
self.n_required = n_required
def two_proportions_sample_size(p1, p2, alpha=0.05, power=0.8, frac=0.5):
ratio = frac / (1. - frac)
es = sms.proportion_effectsize(p1, p2)
n = np.floor(sms.NormalIndPower().solve_power(es,
power=power,
alpha=alpha,
ratio=ratio))
n1, n2 = n * ratio, n
return n1, n2
def post(self):
sig_level = 0.05
power = 0.8
body = request.get_json(silent=True) or {}
p1 = body['p1']
p2 = body['p2']
p1_and_p2 = sms.proportion_effectsize(p1, p2)
sample_size = sms.NormalIndPower().solve_power(p1_and_p2,
power=power,
alpha=sig_level)
return {
'error': False,
'message': 'The required sample size for this campaign is:',
'data': round(sample_size),
'status_code': 200
}, 200
def get_sample_size_for_binomial(p0, p1, power = 0.8, significance = 0.05):
"""
Calculate the sample size for the binomial distribution
Parameters
----------
p0 : [0, 1] current proportion value
p1 : [0, 1] expected proportion value
power : power of the test, e.g. 0.8,
is one minus the probability of a type II error.
Power is the probability that the test correctly
rejects the Null Hypothesis if the Alternative Hypothesis is true.
significance : significance level, e.g. 0.05, is the probability
of a type I error, that is wrong rejections
if the Null Hypothesis is true.
Returns
-------
required_n : Number of observations per sample.
"""
effect_size = sms.proportion_effectsize(p0, p1)
required_n = sms.NormalIndPower().solve_power(effect_size, power=power, alpha=significance)
return ceil(required_n)
NC_cont = payments_cont / clicks_cont
NC_exp = payments_exp / clicks_exp
NC_pooled = (payments_cont + payments_exp) / (clicks_cont + clicks_exp)
NC_sd_pooled = mt.sqrt(NC_pooled * (1 - NC_pooled) *
(1 / clicks_cont + 1 / clicks_exp))
NC_ME = round(get_z_score(1 - alpha / 2) * NC_sd_pooled, 4)
NC_diff = round(NC_exp - NC_cont, 4)
# print("The change due to the experiment is",NC_diff*100,"%")
# print("Confidence Interval: [",NC_diff-NC_ME,",",NC_diff+NC_ME,"]")
# print ("The change is statistically significant if the CI doesn't include 0. In that case, it is practically significant if",NC["d_min"],"is not in the CI as well.")
# Case91_嘗試以函數算出樣本數_Calculating effect size based on our expected rates
effect_size = sms.proportion_effectsize(GC["p"] - 1.0 * GC["d_min"],
GC["p"] + 0.0 * GC["d_min"])
required_n = sms.NormalIndPower().solve_power(effect_size,
power=0.8,
alpha=0.05,
ratio=1)
required_n = ceil(required_n)
print(effect_size, required_n)
# Case02-自行開發雙樣本比例的信賴區間函數
def two_proprotions_confint(success_a,
size_a,
success_b,
size_b,
significance=0.05):
prop_a = success_a / size_a
prop_b = success_b / size_b
var = prop_a * (1 - prop_a) / size_a + prop_b * (1 - prop_b) / size_b
se = np.sqrt(var)
%matplotlib inline
# Some plot styling preferences
plt.style.use('seaborn-whitegrid')
font = {'family' : 'Helvetica',
'weight' : 'bold',
'size' : 14}
mpl.rc('font', **font)
# calculate effect size by propotion
effect_size = sms.proportion_effectsize(0.13, 0.15) # Calculating effect size based on our expected rates
required_n = sms.NormalIndPower().solve_power(
effect_size,
power=0.8,
alpha=0.05,
ratio=1
) # Calculating sample size needed
required_n = ceil(required_n) # Rounding up to next whole number
print(required_n)
# get dataframe info
df.info()
pd.crosstab(df['group'], df['landing_page'])
# get sample from both groups
control_sample = df[df['group'] == 'control'].sample(n=required_n, random_state=22)