def make_anova_context(df, cols, title, fn, cid):
print("\tMAKING ANOVA")
fn = fn + "-" + str(cid)
title_label = CROSS_TITLE[title]
SIGNIFICANCE_CUTOFF = .4
anova_text = title + "\n"
# print("ANOVA FOR ")
# print(analysis_label)
# print(df[analysis_label])
# print(df_col)
# df_col.columns == ['variable', 'value']
# val_min = df_col['value'].get(df_col['value'].idxmin())
# val_max = df_col['value'].get(df_col['value'].idxmax())
# homogenous_data = (val_min == val_max)
homogenous_data = False
# bx = sns.boxplot(x="value", y="question", data=df, palette=custom_palette)
if not homogenous_data:
aov = pg.anova(dv='value', between='question', data=df) #, subject='ResponseId')
aov.round(3)
anova_text = anova_text + str(aov)
aov.to_csv(FILENAME_ANOVAS + fn + '-anova.csv')
p_val = aov['p-unc'][0]
print("\t\t" + title)
print("\t\t" + 'Across contexts:' + "->" + " p=" + str(p_val))
# if p_chair < SIGNIFICANCE_CUTOFF:
# print("Chair position is significant for " + analysis_label + ": " + str(p_chair))
# # print(title)
# if p_path_method < SIGNIFICANCE_CUTOFF:
# print("Pathing method is significant for " + analysis_label + ": " + str(p_path_method))
# # print(title)
# anova_text = anova_text + "\n"
# Verify that subjects is legit
# print(df[subject_id])
posthocs = df.pairwise_ttests(dv='value', between='question', padjust='bonf')
# pg.print_table(posthocs)
anova_text = anova_text + "\n" + str(posthocs)
posthocs.to_csv(FILENAME_ANOVAS + fn + '-posthocs.csv')
else:
print("! Issue creating ANOVA for " + analysis_label)
print("Verify that there are at least a few non-identical values recorded")
anova_text = anova_text + "Column homogenous with value " + str(val_min)
f = open(FILENAME_ANOVAS + fn + "-anova.txt", "w")
f.write(anova_text)
f.close()
def btn_clk11(self):
aov = pg.anova(data=self.df4,
dv='AUC',
between='groupe',
detailed=True).round(6)
aov_res = pd.DataFrame(aov.T).transpose()
model = DataFrameModel(aov_res)
self.tableView.setModel(model)
self.tableView.resizeColumnsToContents()
def anova_eta2(df, ax=None, skip_assert=False):
"""Compute the eta^2 statistic for an ANOVA using the ranking data"""
z = df.stack().reset_index()
z = z.rename(columns={0: 'Rank'})
aov = pg.anova(dv='Rank', between=['Group', 'OdorName'], data=z, detailed=True, effsize='n2').set_index('Source')
if not skip_assert:
assert aov.loc['Group', 'n2'] < 0.0001
aov = aov.drop('Group')
aov.loc['Residual', 'n2'] = 1 - aov['n2'].sum()
aov = aov.rename(index={'Residual': 'Individual', 'Group': 'Culture', 'OdorName': 'Odorant', 'Group * OdorName': 'Culture x Odorant'})
return aov
def aovtables(df, y1, y2):
"""
:param df: dataframe
:param y1: factor 1
:param y2: factor 2
:return: Anova tables (III) testing the influence of y1 and y2 factors on each quantitative variable
"""
aovlist = []
numlist = df.select_dtypes(exclude=['object']).columns.tolist()
for name in numlist:
aovlist.append([
name,
pg.anova(dv=name,
between=[str(y1), str(y2)],
data=df,
detailed=True)
])
return aovlist
def computeAnovas(dv, between_var, data, adjust_type, effect_size_type,
save_dir):
import numpy as np
import pandas as pd
import pingouin as pg
from pingouin import pairwise_ttests
# compute anova
aov = pg.anova(dv=dv, between=between_var, data=data)
aov.to_csv(save_dir + '/' + dv + '_anova.csv', index=False)
print(aov)
# compute pairwise ttests
ttests = pg.pairwise_ttests(dv=dv,
between=between_var,
data=data,
padjust=adjust_type,
effsize=effect_size_type)
ttests.to_csv(save_dir + '/' + dv + '_ttests.csv', index=False)
print(ttests)
def intraclass_corr(data=None, targets=None, raters=None, ratings=None,
nan_policy='raise'):
"""Intraclass correlation.
Parameters
----------
data : :py:class:`pandas.DataFrame`
Long-format dataframe. Data must be fully balanced.
targets : string
Name of column in ``data`` containing the targets.
raters : string
Name of column in ``data`` containing the raters.
ratings : string
Name of column in ``data`` containing the ratings.
nan_policy : str
Defines how to handle when input contains missing values (nan).
`'raise'` (default) throws an error, `'omit'` performs the calculations
after deleting target(s) with one or more missing values (= listwise
deletion).
.. versionadded:: 0.3.0
Returns
-------
stats : :py:class:`pandas.DataFrame`
Output dataframe:
* ``'Type'``: ICC type
* ``'Description'``: description of the ICC
* ``'ICC'``: intraclass correlation
* ``'F'``: F statistic
* ``'df1'``: numerator degree of freedom
* ``'df2'``: denominator degree of freedom
* ``'pval'``: p-value
* ``'CI95%'``: 95% confidence intervals around the ICC
Notes
-----
The intraclass correlation (ICC, [1]_) assesses the reliability of ratings
by comparing the variability of different ratings of the same subject to
the total variation across all ratings and all subjects.
Shrout and Fleiss (1979) [2]_ describe six cases of reliability of ratings
done by :math:`k` raters on :math:`n` targets. Pingouin returns all six
cases with corresponding F and p-values, as well as 95% confidence
intervals.
From the documentation of the ICC function in the `psych
<https://cran.r-project.org/web/packages/psych/psych.pdf>`_ R package:
- **ICC1**: Each target is rated by a different rater and the raters are
selected at random. This is a one-way ANOVA fixed effects model.
- **ICC2**: A random sample of :math:`k` raters rate each target. The
measure is one of absolute agreement in the ratings. ICC1 is sensitive
to differences in means between raters and is a measure of absolute
agreement.
- **ICC3**: A fixed set of :math:`k` raters rate each target. There is no
generalization to a larger population of raters. ICC2 and ICC3 remove
mean differences between raters, but are sensitive to interactions.
The difference between ICC2 and ICC3 is whether raters are seen as fixed
or random effects.
Then, for each of these cases, the reliability can either be estimated for
a single rating or for the average of :math:`k` ratings. The 1 rating case
is equivalent to the average intercorrelation, while the :math:`k` rating
case is equivalent to the Spearman Brown adjusted reliability.
**ICC1k**, **ICC2k**, **ICC3K** reflect the means of :math:`k` raters.
This function has been tested against the ICC function of the R psych
package. Note however that contrarily to the R implementation, the
current implementation does not use linear mixed effect but regular ANOVA,
which means that it only works with complete-case data (no missing values).
References
----------
.. [1] http://www.real-statistics.com/reliability/intraclass-correlation/
.. [2] Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations:
uses in assessing rater reliability. Psychological bulletin, 86(2),
420.
Examples
--------
ICCs of wine quality assessed by 4 judges.
>>> import pingouin as pg
>>> data = pg.read_dataset('icc')
>>> icc = pg.intraclass_corr(data=data, targets='Wine', raters='Judge',
... ratings='Scores').round(3)
>>> icc.set_index("Type")
Description ICC F df1 df2 pval CI95%
Type
ICC1 Single raters absolute 0.728 11.680 7 24 0.0 [0.43, 0.93]
ICC2 Single random raters 0.728 11.787 7 21 0.0 [0.43, 0.93]
ICC3 Single fixed raters 0.729 11.787 7 21 0.0 [0.43, 0.93]
ICC1k Average raters absolute 0.914 11.680 7 24 0.0 [0.75, 0.98]
ICC2k Average random raters 0.914 11.787 7 21 0.0 [0.75, 0.98]
ICC3k Average fixed raters 0.915 11.787 7 21 0.0 [0.75, 0.98]
"""
from pingouin import anova
# Safety check
assert isinstance(data, pd.DataFrame), 'data must be a dataframe.'
assert all([v is not None for v in [targets, raters, ratings]])
assert all([v in data.columns for v in [targets, raters, ratings]])
assert nan_policy in ['omit', 'raise']
# Convert data to wide-format
data = data.pivot_table(index=targets, columns=raters, values=ratings)
# Listwise deletion of missing values
nan_present = data.isna().any().any()
if nan_present:
if nan_policy == 'omit':
data = data.dropna(axis=0, how='any')
else:
raise ValueError("Either missing values are present in data or "
"data are unbalanced. Please remove them "
"manually or use nan_policy='omit'.")
# Back to long-format
# data_wide = data.copy() # Optional, for PCA
data = data.reset_index().melt(id_vars=targets, value_name=ratings)
# Check that ratings is a numeric variable
assert data[ratings].dtype.kind in 'bfiu', 'Ratings must be numeric.'
# Check that data are fully balanced
# This behavior is ensured by the long-to-wide-to-long transformation
# Unbalanced data will result in rows with missing values.
# assert data.groupby(raters)[ratings].count().nunique() == 1
# Extract sizes
k = data[raters].nunique()
n = data[targets].nunique()
# Two-way ANOVA
with np.errstate(invalid='ignore'):
aov = anova(dv=ratings, between=[targets, raters], data=data,
ss_type=2)
# Extract mean squares
msb = aov.at[0, 'MS']
msw = (aov.at[1, 'SS'] + aov.at[2, 'SS']) / (aov.at[1, 'DF'] +
aov.at[2, 'DF'])
msj = aov.at[1, 'MS']
mse = aov.at[2, 'MS']
# Calculate ICCs
icc1 = (msb - msw) / (msb + (k - 1) * msw)
icc2 = (msb - mse) / (msb + (k - 1) * mse + k * (msj - mse) / n)
icc3 = (msb - mse) / (msb + (k - 1) * mse)
icc1k = (msb - msw) / msb
icc2k = (msb - mse) / (msb + (msj - mse) / n)
icc3k = (msb - mse) / msb
# Calculate F, df, and p-values
f1k = msb / msw
df1 = n - 1
df1kd = n * (k - 1)
p1k = f.sf(f1k, df1, df1kd)
f2k = f3k = msb / mse
df2kd = (n - 1) * (k - 1)
p2k = f.sf(f2k, df1, df2kd)
# Create output dataframe
stats = {
'Type': ['ICC1', 'ICC2', 'ICC3', 'ICC1k', 'ICC2k', 'ICC3k'],
'Description': ['Single raters absolute', 'Single random raters',
'Single fixed raters', 'Average raters absolute',
'Average random raters', 'Average fixed raters'],
'ICC': [icc1, icc2, icc3, icc1k, icc2k, icc3k],
'F': [f1k, f2k, f2k, f1k, f2k, f2k],
'df1': n - 1,
'df2': [df1kd, df2kd, df2kd, df1kd, df2kd, df2kd],
'pval': [p1k, p2k, p2k, p1k, p2k, p2k]
}
stats = pd.DataFrame(stats)
# Calculate confidence intervals
alpha = 0.05
# Case 1 and 3
f1l = f1k / f.ppf(1 - alpha / 2, df1, df1kd)
f1u = f1k * f.ppf(1 - alpha / 2, df1kd, df1)
l1 = (f1l - 1) / (f1l + (k - 1))
u1 = (f1u - 1) / (f1u + (k - 1))
f3l = f3k / f.ppf(1 - alpha / 2, df1, df2kd)
f3u = f3k * f.ppf(1 - alpha / 2, df2kd, df1)
l3 = (f3l - 1) / (f3l + (k - 1))
u3 = (f3u - 1) / (f3u + (k - 1))
# Case 2
fj = msj / mse
vn = df2kd * ((k * icc2 * fj + n * (1 + (k - 1) * icc2) - k * icc2))**2
vd = df1 * k**2 * icc2**2 * fj**2 + \
(n * (1 + (k - 1) * icc2) - k * icc2)**2
v = vn / vd
f2u = f.ppf(1 - alpha / 2, n - 1, v)
f2l = f.ppf(1 - alpha / 2, v, n - 1)
l2 = n * (msb - f2u * mse) / (f2u * (k * msj + (k * n - k - n) * mse) +
n * msb)
u2 = n * (f2l * msb - mse) / (k * msj + (k * n - k - n) * mse + n * f2l *
msb)
# Round the confidence intervals
def list_round(x, decimals=2):
for i, xi in enumerate(x):
x[i] = np.round(xi, decimals).tolist()
return x
stats['CI95%'] = list_round([[l1, u1], [l2, u2], [l3, u3],
[1 - 1 / f1l, 1 - 1 / f1u],
[l2 * k / (1 + l2 * (k - 1)),
u2 * k / (1 + u2 * (k - 1))],
[1 - 1 / f3l, 1 - 1 / f3u]])
return stats
def valiate_etaSq(etaSq):
if etaSq < .01:
qual = 'Negligible 微乎其微'
elif etaSq < .06:
qual = 'Small 小有關聯'
elif etaSq < .14:
qual = 'Medium 有關係但影響不大'
else:
qual = 'Large 非常相關'
return qual
dv = "Age"
between = "Survived_cate"
aov = pg.anova(dv=dv, between=between, data=df_train, detailed=True)
etaSq = aov.SS[0] / (aov.SS[0] + aov.SS[1])
print("Q1: 透過數值法計算 Age 和 Survived 是否有相關性? A:連續與離散")
print("Eta Squared (η2)結果：%.3f 相關性 %s" % (etaSq, valiate_etaSq(etaSq)))
# print("Cramer's Values 結果 ", res.loc[2, 'results'], valiate_etaSq(res.loc[2, 'results']))
dt = 'Survived_cate'
between = 'Sex'
# step1: 用交叉列連表(contingency table)，來整理兩個類別型的資料
contTable = pd.crosstab(df_train[between], df_train[dt])
df = min(contTable.shape[0], contTable.shape[1]) - 1
crosstab, res = researchpy.crosstab(df_train[between],
df_train[dt],
test='chi-square')
print()
print("Q2:透過數值法計算 Sex 和 Survived 是否有相關性? A:離散與離散")
print("Cramer's 相關性%.3f 結果 %s" %
if df.iat[i, 7] < 0.042296667:
df.iat[i, 7] = 1
elif df.iat[i,7] < 0.0673333:
df.iat[i, 7] = 2
else:
df.iat[i,7] = 3
df["PageR"].replace({1: "baja", 2: "media", 3: "alta"}, inplace= True)
#logX = np.log10(df['Mediana'])
#df = df.assign(mediana_log=df['Mediana'])
#df.drop(['Mediana'], axis= 1, inplace= True)
factores=["Grado","CoefA","CentCe","CentCa","Excent","PageR"]
plt.figure(figsize=(8, 6))
for i in factores:
anova = pg.anova (dv=df["Flujomaximo"], between=i, data=df, detailed=True )
pg._export_table (anova,("ANOVAs"+i+".csv"))
ax=sns.boxplot(x=df["Flujomaximo"], y=df[i], data=df, palette="cubehelix")
plt.savefig("boxflujo" + i + ".eps", bbox_inches='tight')
tukey = pairwise_tukeyhsd(endog = df["Flujomaximo"], groups= df[i], alpha=0.05)
tukey.plot_simultaneous(xlabel='Flujo máximo', ylabel=i)
plt.vlines(x=49.57,ymin=-0.5,ymax=4.5, color="red")
plt.savefig("simultaneous_flujo" + i + ".eps", bbox_inches='tight')
print(tukey.summary())
t_csv = open("Tukeyflujo"+i+".csv", 'w')
with t_csv:
writer = csv.writer(t_csv)
writer.writerows(tukey.summary())
plt.show()
pca_data = pd.DataFrame(data=comp, columns=['PC1', 'PC2'], index=X.index)
pca_data.index = data_cd56.index
pca_data = pd.merge(pca_data, data, left_index=True,
right_index=True) # merge on both indices
#%%
### STATISTICAL TESTS ###
# For subplots D and D we perform a one-way ANOVA to test the null hypothesis that
# two or more groups have the same population mean. Here all the samples are independent
# as coming from different FCGR3A haplotype and tested only in one condition
# !pip install openpyxl
ANOVA_top = anova(
data=data_adcc,
dv='top', # dependent variable
between='FCGR3A') # between-subject identifier
ANOVA_top.to_excel('../stats/ANOVA_top_figure3.xlsx')
ANOVA_top_posthoc = pairwise_tukey(data=data_adcc, dv='top', between='FCGR3A')
ANOVA_top_posthoc.to_excel('../stats/ANOVA_top_posthoc_figure3.xlsx')
ANOVA_ec50 = anova(
data=data_adcc,
dv='EC50', # dependent variable
between='FCGR3A') # between-subject identifier
ANOVA_ec50.to_excel('../stats/ANOVA_ec50_figure3.xlsx')
ANOVA_ec50_posthoc = pairwise_tukey(data=data_adcc,
dv='EC50',
between='FCGR3A')
file = 'E:/03_FORMATED_DATA/BEHAVIOR/Catwalk_Norm_Profiles_Cuff_Sham_Ctrl.xlsx'
palette = ['royalblue','0.5','lightcoral']
fig, ax = plt.subplots(1,4, sharex=False, sharey=True)
df = pd.read_excel(file, header=0)
#Easy first, peak amplitude at D15
postOp15 = df[['post_op_15','Condition']]
sn.boxplot(x='Condition',y='post_op_15',data=df,ax=ax[0],palette=['lightcoral','0.5','royalblue'])
#sn.swarmplot(x='Condition',y='post_op_15',data=postOp15,ax=ax[0],color='black')
#[1] --------------------Stats on PostOp15 peak-------------------------------------
postOp15_KW = pg.kruskal(data=postOp15,dv='post_op_15',between='Condition')
postOp15_Anova = pg.anova(data=postOp15,dv='post_op_15',between='Condition')
print('Analysis of behavioral features')
print ('[1]--------------------- Post Op 15 ------------------------')
print('Average values')
print(postOp15.groupby('Condition').mean())
print('')
print('STD')
print(postOp15.groupby('Condition').std())
print('')
print('Multi condition test')
print ('Kruskal Wallis')
print (postOp15_KW)
if postOp15_KW['p-unc'].values <= 0.05:
print ('Post Hoc MWU tests')
print('Percentiles: 25th=%.3f, 75th=%.3f, IQR=%.3f' % (q25, q75, iqr))
# calculate the outlier cutoff
cut_off = iqr * 1.5
print(cut_off)
lower, upper = q25 - cut_off, q75 + cut_off
# identify outliers
outliers = [x for x in dataset['rssi'] if x < lower or x > upper]
print('Identified outliers: %d' % len(outliers))
# remove outliers
outliers_removed = [x for x in dataset['rssi'] if x >= lower and x <= upper]
print('Non-outlier observations: %d' % len(outliers_removed))
#two-way Annova test
#P value of the tag>significant level reject null hypothesis
aov = pg.anova(dv='rssi', between=['Tag', 'Antenna'], data=dataset,
detailed=True)
print(aov)
#number =LabelEncoder()
#dataset['Tag'] = number.fit_transform(dataset.Tag)
jobs_encoder = LabelBinarizer()
jobs_encoder.fit(dataset['Tag'])
transformed = jobs_encoder.transform(dataset['Tag'])
ohe_df = pd.DataFrame(transformed)
dataset = pd.concat([dataset, ohe_df], axis=1).drop(['Tag'], axis=1)
X = dataset.iloc[:,4:].values
y = dataset.iloc[:,3].values
#X=np.reshape(X, (202004, 1))
#y=np.reshape(y,(202004, 1))
def intraclass_corr(data=None, items=None, raters=None, scores=None, ci=.95):
"""Intra-class correlation coefficient.
Parameters
----------
data : pd.DataFrame
Dataframe containing the variables
items : string
Name of column in data containing the items (targets).
raters : string
Name of column in data containing the raters (scorers).
scores : string
Name of column in data containing the scores (ratings).
ci : float
Confidence interval
Returns
-------
icc : float
Intraclass correlation coefficient
ci : list
Lower and upper confidence intervals
Notes
-----
The intraclass correlation (ICC) assesses the reliability of ratings by
comparing the variability of different ratings of the same subject to the
total variation across all ratings and all subjects. The ratings are
quantitative (e.g. Likert scale).
Shrout and Fleiss (1979) describe six cases of reliability of ratings done
by :math:`k` raters on :math:`n` targets. Pingouin only returns ICC1,
which consider that each target is rated by a different rater and the
raters are selected at random. (This is a one-way ANOVA
fixed effects model and is found by (MSB - MSW)/(MSB + (nr - 1) * MSW)).
ICC1 is sensitive to differences in means between raters and is a measure
of absolute agreement.
This function has been tested against the ICC function of the R psych
package.
References
----------
.. [1] Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations:
uses in assessing rater reliability. Psychological bulletin, 86(2),
420.
.. [2] https://cran.r-project.org/web/packages/psych/psych.pdf
.. [3] http://www.real-statistics.com/reliability/intraclass-correlation/
Examples
--------
ICC of wine quality assessed by 4 judges.
>>> import pingouin as pg
>>> data = pg.read_dataset('icc')
>>> pg.intraclass_corr(data=data, items='Wine', raters='Judge',
... scores='Scores', ci=.95)
(0.727526, array([0.434, 0.927]))
"""
from pingouin import anova
# Check dataframe
if any(v is None for v in [data, items, raters, scores]):
raise ValueError('Data, items, raters and scores must be specified')
assert isinstance(data, pd.DataFrame), 'Data must be a pandas dataframe.'
# Check that scores is a numeric variable
assert data[scores].dtype.kind in 'fi', 'Scores must be numeric.'
# Check that data are fully balanced
if data.groupby(raters)[scores].count().nunique() > 1:
raise ValueError('Data must be balanced.')
# Extract sizes
k = data[raters].nunique()
# n = data[groups].nunique()
# ANOVA and ICC
aov = anova(dv=scores, data=data, between=items, detailed=True)
icc = (aov.at[0, 'MS'] - aov.at[1, 'MS']) / \
(aov.at[0, 'MS'] + (k - 1) * aov.at[1, 'MS'])
# Confidence interval
alpha = 1 - ci
df_num, df_den = aov.at[0, 'DF'], aov.at[1, 'DF']
f_lower = aov.at[0, 'F'] / f.isf(alpha / 2, df_num, df_den)
f_upper = aov.at[0, 'F'] * f.isf(alpha / 2, df_den, df_num)
lower = (f_lower - 1) / (f_lower + k - 1)
upper = (f_upper - 1) / (f_upper + k - 1)
return round(icc, 6), np.round([lower, upper], 3)
pk1 = open('./stats_df_1.pkl', 'rb')
pk2 = open('./stats_df_2.pkl', 'rb')
df1 = pickle.load(pk1)
df2 = pickle.load(pk2)
# DF1
# PLOT
plt.subplot(121)
seaborn.stripplot(y=df1['RMSE'],
x=df1['co_contraction_level'],
hue=df1['EMG_objective'])
# STATS
aov = pg.anova(dv='RMSE',
between=['EMG_objective', 'co_contraction_level'],
data=df1)
ptt = pg.pairwise_ttests(dv='RMSE',
between=['EMG_objective', 'co_contraction_level'],
data=df1,
padjust='bonf')
pg.print_table(aov.round(3))
pg.print_table(ptt.round(3))
# DF2
# PLOT
plt.subplot(122)
seaborn.stripplot(y=df2['RMSE'],
x=df2['Marker_noise_level_m'],
hue=df2['EMG_objective'])
aov = pg.anova(dv='RMSE',
y = bins % 2
dataset = pd.DataFrame({'x1': x1, 'x2': x2, 'y': y})
display(dataset)
# %%
dataset.info()
# %%
columns = dataset.columns.tolist()
target = ['y']
feature_names = [item for item in columns if item not in target]
rows = len(dataset)
print('All columns: ' + str(columns))
print('Feature names: ' + str(feature_names))
print('Target: ' + str(target))
print('Number of rows: ' + str(rows))
# %%
pg.anova(data=dataset, dv='x1', between='y', detailed=True)
# %%
pg.anova(data=dataset, dv='x2', between='y', detailed=True)
# %%
X = dataset[['x1']]
y = dataset[target]
rows = len(X)
min_samples_leaf = int(rows * 0.05)
classifier = DecisionTreeClassifier(criterion='gini',
min_samples_leaf=min_samples_leaf)
classifier.fit(X, y)
# %%
X_columns = X.columns.tolist()
text_representation = export_text(classifier, feature_names=X_columns)
print(text_representation)
# %%
def stats(model, quantity, data, targets, tw, rm, nd):
if model == 'absolute':
data = data.drop(['NormQuant'], axis=1)
data['NormMean'] = data['NormMean'].astype(float)
mean = 'NormMean'
else:
data = data.drop(['rq'], axis=1)
data['rqMean'] = data['rqMean'].astype(float)
mean = 'rqMean'
# prepare data from intermediate dataframe
data = data[data['Outliers'].eq(False)]
data = data.drop_duplicates(keep='first')
# t-test and anova for normally distributed data
if nd == 'True':
if quantity == 2:
# T-Test between 2 groups
stats_dfs = pandas.DataFrame()
posthoc_dfs = pandas.DataFrame()
group = data['Group'].dropna()
group = group.drop_duplicates(keep='first').values.tolist()
for item in targets:
df = data[data['Target Name'].eq(item)]
group1 = df[df['Group'].eq(group[0])][mean]
group2 = df[df['Group'].eq(group[1])][mean]
t_test = ttest(group1, group2, paired=bool(rm))
if rm == 'True':
t_test['paired'] = 'TRUE'
else:
t_test['paired'] = 'FALSE'
t_test['Target Name'] = item
if stats_dfs is None:
stats_dfs = t_test
else:
stats_dfs = stats_dfs.append(t_test, ignore_index=True)
# reformat output table
stats_dfs = stats_dfs.rename(columns={
'cohen-d': 'effect size',
'BF10': 'Bayes factor',
'dof': 'DF'
})
cols = [
'Target Name', 'DF', 'T', 'tail', 'paired', 'p-val',
'effect size', 'power', 'Bayes factor'
]
stats_dfs = stats_dfs.reindex(columns=cols)
elif quantity >= 3:
# ANOVA test
stats_dfs = pandas.DataFrame()
posthoc_dfs = pandas.DataFrame()
# tukey_dfs = pandas.DataFrame()
pvals = []
for item in targets:
if rm == 'True':
# one-way
if tw == 'False':
# repeated measure anova
aov = pg.rm_anova(
dv=mean,
data=data[data['Target Name'].eq(item)],
within='Group',
subject='Sample Name',
detailed=True)
pvals.append(aov['p-unc'][0])
aov = aov.drop([1])
aov['measures'] = ['dependent']
aov['Target Name'] = item
# two-way
else:
aov = pg.rm_anova(
dv=mean,
data=data[data['Target Name'].eq(item)],
within=['Group1', 'Group2'],
subject='Sample Name',
detailed=True)
reject_tw, pval_corr_tw = pg.multicomp(list(
aov['p-unc']),
alpha=0.05,
method='bonf')
aov['p-value corrected'] = pval_corr_tw
aov['measures'] = ['dependent'] * 3
aov['Target Name'] = [item] * 3
aov.drop(['eps'], axis=1)
ph = pairwise_ttests(
data=data[data['Target Name'].eq(item)],
dv=mean,
within='Group',
subject='Sample Name',
padjust='fdr_bh')
ph['Target Name'] = item
ph['Test'] = 'T-Test'
else:
# one-way
if tw == 'False':
aov = pg.anova(dv=mean,
between='Group',
data=data[data['Target Name'].eq(item)],
detailed=True)
pvals.append(aov['p-unc'][0])
aov = aov.drop([1])
aov['measures'] = ['independent']
aov['Target Name'] = item
ph = pairwise_ttests(
data=data[data['Target Name'].eq(item)],
dv=mean,
between='Group',
padjust='fdr_bh')
ph['Test'] = 'T-Test'
# two-way
else:
aov = pg.anova(dv=mean,
between=['Group1', 'Group2'],
data=data[data['Target Name'].eq(item)],
detailed=False)
aov = aov.drop([3])
reject_tw, pval_corr_tw = pg.multicomp(list(
aov['p-unc']),
alpha=0.05,
method='bonf')
aov['p-value corrected'] = pval_corr_tw
aov['measures'] = ['independent'] * 3
aov['Target Name'] = [item] * 3
ph = pairwise_ttests(
data=data[data['Target Name'].eq(item)],
dv=mean,
between=['Group1', 'Group2'],
padjust='fdr_bh')
ph['Test'] = 'T-Test'
ph['Target Name'] = item
if stats_dfs is None:
stats_dfs = aov
else:
stats_dfs = stats_dfs.append(aov, ignore_index=True)
if posthoc_dfs is None:
posthoc_dfs = ph
else:
posthoc_dfs = posthoc_dfs.append(ph, ignore_index=True)
reject, pvals_corr = pg.multicomp(pvals, alpha=0.05, method='bonf')
# reformat output tables
stats_dfs = stats_dfs.rename(columns={
'p-unc': 'p-value',
'np2': 'effect size'
})
if tw == 'False':
stats_dfs['p-value corrected'] = pvals_corr
stats_dfs['distribution'] = ['parametric'] * len(targets)
stats_dfs['test'] = ['ANOVA'] * len(targets)
stats_dfs['statistic'] = ['NA'] * len(targets)
else:
stats_dfs['distribution'] = ['parametric'] * (len(targets) * 3)
stats_dfs['test'] = ['ANOVA'] * (len(targets) * 3)
stats_dfs['statistic'] = ['NA'] * (len(targets) * 3)
cols = [
'Target Name', 'Source', 'DF', 'F', 'MS', 'SS', 'p-value',
'p-value corrected', 'measures', 'distribution', 'test',
'statistic', 'effect size'
]
stats_dfs = stats_dfs.reindex(columns=cols)
if tw == 'False':
posthoc_dfs = posthoc_dfs.drop(['Contrast', 'T'], axis=1)
else:
posthoc_dfs = posthoc_dfs.drop(['T'], axis=1)
posthoc_dfs = posthoc_dfs.rename(
columns={
'hedges': 'effect size',
'p-corr': 'p-value corrected',
'p-unc': 'p-value',
'p-adjust': 'correction method',
'BF10': 'Bayes factor',
'dof': 'DF'
})
if tw == 'False':
cols2 = [
'Target Name', 'A', 'B', 'DF', 'p-value corrected',
'p-value', 'correction method', 'Paired', 'Parametric',
'Test', 'effect size', 'Bayes factor'
]
else:
cols2 = [
'Target Name', 'Contrast', 'Group1', 'A', 'B', 'DF',
'p-value corrected', 'p-value', 'correction method',
'Paired', 'Parametric', 'Test', 'effect size',
'Bayes factor'
]
posthoc_dfs = posthoc_dfs.reindex(columns=cols2)
# nonparametric tests for not normally distributed data
else:
if quantity == 2:
stats_dfs = pandas.DataFrame()
posthoc_dfs = pandas.DataFrame()
group = data['Group'].dropna()
group = group.drop_duplicates(keep='first').values.tolist()
for item in targets:
df = data[data['Target Name'].eq(item)]
group1 = df[df['Group'].eq(group[0])][mean]
group2 = df[df['Group'].eq(group[1])][mean]
if rm == 'True':
# Mann-Whitney U test
test = mannwhitneyu(group1, group2)
test = pandas.DataFrame(
{
'Target Name': item,
'pvalue': test.pvalue,
'statistic': test.statistic
},
index=[0])
else:
# Wilcoxon
test = wilcoxon(group1, group2)
test = pandas.DataFrame(
{
'Target Name': item,
'pvalue': test.pvalue,
'statistic': test.statistic
},
index=[0])
if stats_dfs is None:
stats_dfs = test
else:
stats_dfs = stats_dfs.append(test, ignore_index=True)
elif quantity >= 3:
stats_dfs = pandas.DataFrame()
posthoc_dfs = pandas.DataFrame()
pvals = []
for item in targets:
if rm == 'True':
# friedman test for repeated measurements
df = pg.friedman(dv=mean,
within='Group',
subject='Sample Name',
data=data[data['Target Name'].eq(item)])
pvals.append(df['p-unc'][0])
df['test'] = ['Friedman Q']
df['measures'] = ['dependent']
df = df.rename(columns={'Q': 'statistic'})
df['Target Name'] = item
df['DF'] = 'NA'
ph = pairwise_ttests(
data=data[data['Target Name'].eq(item)],
dv=mean,
within='Group',
subject='Sample Name',
padjust='fdr_bh',
parametric=False)
ph['Target Name'] = item
ph['DF'] = 'NA'
ph['Bayes factor'] = 'NA'
ph['Test'] = 'Wilcoxon'
else:
# Kruskal-Wallis H test
df = pg.kruskal(dv=mean,
between='Group',
data=data[data['Target Name'].eq(item)])
pvals.append(df['p-unc'][0])
df['test'] = ['Kruskal-Wallis H']
df['measures'] = ['independent']
df = df.rename(columns={'H': 'statistic'})
df['Target Name'] = item
df['DF'] = 'NA'
ph = pairwise_ttests(
data=data[data['Target Name'].eq(item)],
dv=mean,
between='Group',
padjust='fdr_bh',
parametric=False)
ph['Target Name'] = item
ph['DF'] = 'NA'
ph['Bayes factor'] = 'NA'
ph['Test'] = 'Mann-Whitney U'
if stats_dfs is None:
stats_dfs = df
else:
stats_dfs = stats_dfs.append(df, ignore_index=True)
if posthoc_dfs is None:
posthoc_dfs = ph
else:
posthoc_dfs = posthoc_dfs.append(ph, ignore_index=True)
reject, pvals_corr = pg.multicomp(pvals, alpha=0.05, method='bonf')
# reformat output tables
stats_dfs = stats_dfs.rename(columns={
'dof': 'DF',
'p-unc': 'p-value'
})
stats_dfs['p-value corrected'] = pvals_corr
stats_dfs['distribution'] = ['non-parametric'] * len(targets)
stats_dfs['MS'] = ['NA'] * len(targets)
stats_dfs['SS'] = ['NA'] * len(targets)
stats_dfs['effect size'] = ['NA'] * len(targets)
cols = [
'Target Name', 'DF', 'MS', 'SS', 'p-value',
'p-value corrected', 'measures', 'distribution', 'test',
'statistic', 'effect size'
]
stats_dfs = stats_dfs.reindex(columns=cols)
posthoc_dfs = posthoc_dfs.drop(['Contrast'], axis=1)
posthoc_dfs = posthoc_dfs.rename(
columns={
'hedges': 'effect size',
'p-corr': 'p-value corrected',
'p-unc': 'p-value',
'p-adjust': 'correction method',
'BF10': 'Bayes factor'
})
cols2 = [
'Target Name', 'A', 'B', 'DF', 'p-value corrected', 'p-value',
'correction method', 'Paired', 'Parametric', 'Test',
'effect size', 'Bayes factor'
]
posthoc_dfs = posthoc_dfs.reindex(columns=cols2)
return stats_dfs, posthoc_dfs
def test_pandas(self):
"""Test pandas method.
"""
# Test the ANOVA (Pandas)
aov = df.anova(dv='Scores', between='Group', detailed=True)
assert aov.equals(
pg.anova(dv='Scores', between='Group', detailed=True, data=df))
# Test the Welch ANOVA (Pandas)
aov = df.welch_anova(dv='Scores', between='Group')
assert aov.equals(pg.welch_anova(dv='Scores', between='Group',
data=df))
# Test the repeated measures ANOVA (Pandas)
aov = df.rm_anova(dv='Scores',
within='Time',
subject='Subject',
detailed=True)
assert aov.equals(
pg.rm_anova(dv='Scores',
within='Time',
subject='Subject',
detailed=True,
data=df))
# FDR-corrected post hocs with Hedges'g effect size
ttests = df.pairwise_ttests(dv='Scores',
within='Time',
subject='Subject',
padjust='fdr_bh',
effsize='hedges')
assert ttests.equals(
pg.pairwise_ttests(dv='Scores',
within='Time',
subject='Subject',
padjust='fdr_bh',
effsize='hedges',
data=df))
# Test two-way mixed ANOVA
aov = df.mixed_anova(dv='Scores',
between='Group',
within='Time',
subject='Subject',
correction=False)
assert aov.equals(
pg.mixed_anova(dv='Scores',
between='Group',
within='Time',
subject='Subject',
correction=False,
data=df))
# Test parwise correlations
corrs = data.pairwise_corr(columns=['X', 'M', 'Y'], method='spearman')
corrs2 = pg.pairwise_corr(data=data,
columns=['X', 'M', 'Y'],
method='spearman')
assert corrs['r'].equals(corrs2['r'])
# Test partial correlation
corrs = data.partial_corr(x='X', y='Y', covar='M', method='spearman')
corrs2 = pg.partial_corr(x='X',
y='Y',
covar='M',
method='spearman',
data=data)
assert corrs['r'].equals(corrs2['r'])
# Test partial correlation matrix (compare with the ppcor package)
corrs = data.pcorr().round(3)
np.testing.assert_array_equal(corrs.iloc[0, :].values,
[1, 0.392, 0.06, -0.014, -0.149])
# Now compare against Pingouin's own partial_corr function
corrs = data[['X', 'Y', 'M']].pcorr()
corrs2 = data.partial_corr(x='X', y='Y', covar='M')
assert round(corrs.loc['X', 'Y'], 3) == corrs2.loc['pearson', 'r']
# Test mediation analysis
med = data.mediation_analysis(x='X', m='M', y='Y', seed=42, n_boot=500)
np.testing.assert_array_equal(med.loc[:, 'coef'].values,
[0.5610, 0.6542, 0.3961, 0.0396, 0.3565])
import numpy as np
import pandas as pd
import seaborn as sns
import statsmodels
from pingouin import mixed_anova, anova, pairwise_tukey
from pingouin import logistic_regression
data_merged = pd.read_csv(
r'C:\Users\user\Desktop\FOCUS\behavioral\P_Merged_var.csv')
### Fill in Nan values in false alarm and omission error (0)
data_merged = data_merged.fillna({'false_alarm': 0, 'om_err': 0})
#data_merged.to_csv(r'C:\Users\user\Desktop\FOCUS\behavioral\P_Merged_var.csv', index = None, header=True)
# ANOVA - does correct reaction time differ between blocks?
aov_corr_rt = anova(dv='corr_rt', between='blocks', data=data_merged)
print(aov_corr_rt)
rep_anov_alarm = pg.rm_anova(data=data_merged,
dv='false_alarm',
within='blocks',
subject='participant',
detailed=True)
# follow-up pairwise comparison
pairs_corr_rt = pairwise_tukey(dv='corr_rt',
between='blocks',
data=data_merged)
print(pairs_corr_rt)
data = pd.DataFrame(data=dict_dat)
### Store data as csv
data.to_csv(r'data.csv')
### Plot_Data
sns.kdeplot(data=one)
sns.kdeplot(data=two)
sns.kdeplot(data=three)
sns.kdeplot(data=np.append(one, [two, three]))
res = pg.anova(data=data,
dv='data',
between='group',
ss_type=2,
detailed=True,
effsize='np2')
### ANOVA
m_1, var_1, sd_1 = np.mean(one), np.var(one), np.std(one)
m_2, var_2, sd_2 = np.mean(two), np.var(two), np.std(two)
m_3, var_3, sd_3 = np.mean(three), np.var(three), np.std(three)
m_tot, var_tot = data['data'].mean(), data['data'].var()
n_obs = len(one)
sq_tot = ((dict_dat['data'] - dict_dat['data'].mean())**2).sum()
var_dep = sq_tot / (len(dict_dat['data']))
sqa = 35 * np.sum((m_tot - m_1)**2) + 35 * np.sum(
(m_tot - m_2)**2) + 35 * np.sum((m_tot - m_3)**2)
sqr = (n_obs - 1) * var_1 + (n_obs - 1) * var_2 + (n_obs - 1) * var_3
for var_ in vars_time:
print(f'Normality test (Shapiro), {var_}')
print(pg.normality(test_df, dv=var_, group='Group'))
# Within group rm ANOVA to determine side preference
#Prep df with only Sample and Test phases
rm_df = si_raw[si_raw['Phase'] != 'Habituation']
rm_df = pd.melt(
rm_df,
id_vars=['Subject', 'Group', 'Phase'],
value_vars=['Time Object/New Cons Chamber', 'Time Conspecific Chamber'],
var_name='Side',
value_name='Time')
# Run ANOVA
samp_anova = pg.anova(data=rm_df[rm_df['Phase'] == 'Sample'],
dv='Time',
between=['Side', 'Group'])
#Save to csv
samp_anova.to_csv(
'/Users/labc02/Documents/PDCB_data/Behavior/Stats/sample_time_anova.csv')
# post hoc test, pairwise_ttests, holm-bonf correction
samp_posthoc = pg.pairwise_ttests(data=rm_df[rm_df['Phase'] == 'Sample'],
dv='Time',
between=['Group', 'Side'],
padjust='holm')
samp_posthoc.to_csv(
'/Users/labc02/Documents/PDCB_data/Behavior/Stats/sample_time_posthoc.csv')
samp_posthoc[['Group', 'p-corr']]
ph_val = zip(['***', '***', '***', '*'], [6, 55, 102, 155],
[120, 100, 120, 120])
for ii, jj, kk in ph_val:
from statsmodels.formula.api import ols
import numpy as np
import pingouin as pg
import seaborn as sns
from statsmodels.stats.multicomp import pairwise_tukeyhsd
import csv
df = pd.read_csv("Datos.csv", index_col=None,usecols=[1,2,3,4,8],dtype={'generador': 'category',
'algoritmo_flujo': 'category','vertices': 'category','aristas': 'category', 'mediana': np.float64} )
logX = np.log1p(df['mediana'])
df = df.assign(mediana_log=logX.values)
df.drop(['mediana'], axis= 1, inplace= True)
factores=["vertices","generador","aristas","algoritmo_flujo"]
plt.figure(figsize=(8, 6))
for i in factores:
print(rp.summary_cont(df['mediana_log'].groupby(df[i])))
anova = pg.anova (dv='mediana_log', between=i, data=df, detailed=True)
pg._export_table (anova,("ANOVA"+i+".csv"))
ax=sns.boxplot(x=df["mediana_log"], y=df[i], data=df, palette="Set1")
plt.savefig("boxplot_"+ i+".png", bbox_inches='tight')
plt.savefig("boxplot_" + i + ".eps", bbox_inches='tight')
tukey = pairwise_tukeyhsd(endog = df["mediana_log"], groups= df[i], alpha=0.05)
tukey.plot_simultaneous(xlabel='Time', ylabel=i)
plt.vlines(x=49.57,ymin=-0.5,ymax=4.5, color="red")
plt.savefig("simultaneous_tukey"+ i+".png", bbox_inches='tight')
plt.savefig("simultaneous_tukey" + i + ".eps", bbox_inches='tight')
print(tukey.summary())
t_csv = open("Tukey"+i+".csv", 'w')
with t_csv:
writer = csv.writer(t_csv)
writer.writerows(tukey.summary())
plt.show()
# Q1. Does the condition affect the marker value?
# One way ANOVA Value ~ Protocol with PH value | protocol
marker_test_dir = save_test_dir / "01_markers"
if not os.path.exists(marker_test_dir):
os.mkdir(marker_test_dir)
marker_ph_dict = {}
for marker_label, marker_df in zip(marker_dict.keys(), marker_dict.values()):
print(marker_label)
# run anova
curr_anova_marker = pg.anova(
dv=dep_var,
between=condition_col,
data=marker_df
)
pg.print_table(curr_anova_marker)
curr_ph_marker = pg.pairwise_tukey(
dv=dep_var,
between=condition_col,
data=marker_df
)
pg.print_table(curr_ph_marker)
marker_ph_dict[marker_label] = curr_ph_marker
# save the files
label_test_dir = marker_test_dir / marker_label
if not os.path.exists(label_test_dir):
for year, df in metrics.groupby(['Year']):
for level, data in df.groupby('type_cat'):
pubs_list[f'{year}_{level}'] = list(data['pubs_awarded'])
fwci_list[f'{year}_{level}'] = list(data['fwci_awarded'])
# Generate separate dataframes
pubs = pd.DataFrame(dict([(k, pd.Series(v)) for k, v in pubs_list.items()]))
fwci = pd.DataFrame(dict([(k, pd.Series(v)) for k, v in fwci_list.items()]))
FileHandling.df_to_excel(data_frames=[pubs, fwci], sheetnames=['pubs', 'fwci'], output_path=f'{output_folder}metrics_per_year.xlsx')
# Collect cols for each level
levels = ['_1', '_2', '_3']
for level in levels:
cols = [col for col in pubs.columns.tolist() if level in col]
test_df = pubs[cols].melt(value_name='publications', var_name='Group')
# Two-way ANOVA
aov = pg.anova(data=test_df, dv='publications', between='Group',
export_filename=f'{output_folder}anova_pubs{level}.csv')
pg.print_table(aov)
# FDR-corrected post hocs with Hedges'g effect size
posthoc = pg.pairwise_ttests(data=test_df, dv='publications', between='Group', within=None, parametric=True, alpha=.05, tail='two-sided', padjust='bonf', effsize='none', return_desc=False, export_filename=f'{output_folder}bonf_pubs{level}.csv')
# Pretty printing of table
pg.print_table(posthoc, floatfmt='.3f')
print(df["Excentricidad"])
logX = np.log1p(df['Mediana'])
df = df.assign(mediana_log=logX.values)
df.drop(['Mediana'], axis=1, inplace=True)
factores = [
"Grado", "CoefAg", "CentCer", "CentCag", "Excentricidad", "PageRag"
]
plt.figure(figsize=(8, 6))
for i in factores:
print(rp.summary_cont(df['FlujoMax'].groupby(df[i])))
anova = pg.anova(
dv='FlujoMax',
between=i,
data=df,
detailed=True,
)
pg._export_table(anova, ("ANOVAsFlujoMax" + i + ".csv"))
ax = sns.boxplot(x=df["FlujoMax"], y=df[i], data=df, palette="cubehelix")
plt.savefig("boxplot_FlujoMax" + i + ".eps", bbox_inches='tight')
tukey = pairwise_tukeyhsd(endog=df["FlujoMax"], groups=df[i], alpha=0.05)
tukey.plot_simultaneous(xlabel='Flujo Maximo', ylabel=i)
plt.savefig("simultaneous_tukey" + i + ".eps", bbox_inches='tight')
print(tukey.summary())
t_csv = open("TukeyFlujoMax" + i + ".csv", 'w')
def intraclass_corr(data=None, groups=None, raters=None, scores=None, ci=.95):
"""Intra-class correlation coefficient.
Parameters
----------
data : pd.DataFrame
Dataframe containing the variables
groups : string
Name of column in data containing the groups.
raters : string
Name of column in data containing the raters (scorers).
scores : string
Name of column in data containing the scores (ratings).
ci : float
Confidence interval
Returns
-------
icc : float
Intraclass correlation coefficient
ci : list
Lower and upper confidence intervals
Notes
-----
The intraclass correlation (ICC) assesses the reliability of ratings by
comparing the variability of different ratings of the same subject to the
total variation across all ratings and all subjects. The ratings are
quantitative (e.g. Likert scale).
Translated in Python from:
http://www.real-statistics.com/reliability/intraclass-correlation/
Examples
--------
1. ICC of wine quality assessed by 4 judges.
>>> from pingouin.datasets import read_dataset
>>> from pingouin import intraclass_corr
>>> data = read_dataset('icc')
>>> intraclass_corr(data, 'Wine', 'Judge', 'Scores')
(0.727525596259691, array([0.434, 0.927]))
"""
from pingouin import anova
from scipy.stats import f
# Check dataframe
if any(v is None for v in [data, groups, raters, scores]):
raise ValueError('Data, groups, raters and scores must be specified')
if not isinstance(data, pd.DataFrame):
raise ValueError('Data must be a pandas dataframe.')
# Check that scores is a numeric variable
if data[scores].dtype.kind not in 'fi':
raise ValueError('Scores must be numeric.')
# Check that data are fully balanced
if data.groupby(raters)[scores].count().unique().size > 1:
raise ValueError('Data must be balanced.')
# Extract sizes
k = data[raters].unique().size
# n = data[groups].unique().size
# ANOVA and ICC
aov = anova(dv=scores, data=data, between=groups, detailed=True)
icc = (aov.loc[0, 'MS'] - aov.loc[1, 'MS']) / \
(aov.loc[0, 'MS'] + (k - 1) * aov.loc[1, 'MS'])
# Confidence interval
alpha = 1 - ci
df_num, df_den = aov.loc[0, 'DF'], aov.loc[1, 'DF']
f_lower = aov.loc[0, 'F'] / f.isf(alpha / 2, df_num, df_den)
f_upper = aov.loc[0, 'F'] * f.isf(alpha / 2, df_den, df_num)
lower = (f_lower - 1) / (f_lower + k - 1)
upper = (f_upper - 1) / (f_upper + k - 1)
return icc, np.round([lower, upper], 3)
# Statistical analysis on excitations with an Anova method
df_stats = pd.DataFrame({
"RMSE": RMSEtrack,
"co_contraction_level": co_lvl_df,
"EMG_objective": EMG_n_lvl_df,
"Marker_noise_level_m": marker_n_lvl_df,
"component": states_controls_df,
"weight_level": weight_lvl_df,
})
df_stats = df_stats[df_stats["component"] == "force"]
df_stats = df_stats[df_stats["weight_level"] == "high"]
df_stats = df_stats[df_stats["RMSE"].notna()]
df_stats.to_pickle("stats_df_1.pkl")
aov = pg.anova(dv="RMSE",
between=["EMG_objective", "co_contraction_level"],
data=df_stats)
ptt = pg.pairwise_ttests(
dv="RMSE",
between=[
"co_contraction_level",
"EMG_objective",
],
data=df_stats,
padjust="bonf",
)
pg.print_table(aov.round(3))
pg.print_table(ptt.round(3))
# Figure of RMSE on force function of co-contraction level (Fig. 7)
import matplotlib
== "Omnibus test of normality" else "shapiro")
if y_var2 == "None":
st.write(normality)
if y_var2 == "None":
st.success("Levene test for homoscedasticity of variances")
homoscedasticity = pg.homoscedasticity(df, dv=x_var, group=y_var)
if y_var2 == "None":
st.write(homoscedasticity)
if classic_vs_welch == "Classic ANOVA":
if y_var2 == "None":
anova = pg.anova(dv=x_var, between=y_var, data=df, detailed=True)
st.success("Classic ANOVA results")
else:
anova = df.anova(dv=x_var, between=[y_var, y_var2])
st.success("Two-way ANOVA results")
st.write(anova.round(3))
if y_var2 == "None":
st.success("Tukey HSD multiple comparisons")
tukey_mult = df.pairwise_tukey(dv=x_var, between=y_var).round(3)
st.write(tukey_mult)
if y_var2 != "None":
st.success("Tukey HSD multiple comparisons")
Q_data_raw = pd.read_csv(r'C:\Users\de_hauk\PowerFolders\EXPRA_Peper_SS_2020\Daten\fragebogen\fb_exp.txt', sep='\t').sort_values('ID').set_index('ID').drop(labels='group', axis =1 )
#BSI_dat_fin = pd.concat([BSI_dat, Q_data_raw], axis=0)
BSI_dat_fin = pd.merge(BSI_dat, Q_data_raw, left_index=True, right_index=True)
#### Save Data
#BSI
BSI_dat_fin.to_csv(r'C:\Users\de_hauk\PowerFolders\EXPRA_Peper_SS_2020\Daten\data_JASP_2_BSI.csv', sep=',' )
# trial_type/ 2-Faktor
anova_2f_bs = AAA_dat.copy()[['id','RT_M_bl','trial_type','group']]
anova_2f_bs.to_csv(r'C:\Users\de_hauk\PowerFolders\EXPRA_Peper_SS_2020\Daten\2f_anova\anova_2f_bs.csv', sep=',' )
aov_2F = pg.anova(dv='BSI_RT_BL', between='group', data=BSI_dat,
detailed=True)
#### Plots
sns_plot_BSI = sns.catplot(x="group", y="BSI_RT_BL", data=BSI_dat, kind = 'box')
sns_plot_BSI.savefig(r"C:\Users\de_hauk\PowerFolders\EXPRA_Peper_SS_2020\Daten\2f_anova\1F_dat_BSI.png")
plt.clf()
sns_plot_tt = sns.catplot(x='trial_type', y="RT_M_bl", data=AAA_dat, hue = "group", kind = 'box')
sns_plot_tt.savefig(r"C:\Users\de_hauk\PowerFolders\EXPRA_Peper_SS_2020\Daten\2f_anova\2F_dat_tt.png")
anova_2F1 = pg.anova(dv='RT_M_bl', between=['group','trial_type'], data=AAA_dat,detailed=True)
anova_2F1_ss_tot = ((AAA_dat['RT_M_bl']-AAA_dat['RT_M_bl'].mean())**2).sum()
anova_2F1_np_2_group = 2365/(2365 + 308246)
def test_pandas(self):
"""Test pandas method.
"""
# Test the ANOVA (Pandas)
aov = df.anova(dv='Scores', between='Group', detailed=True)
assert aov.equals(
pg.anova(dv='Scores', between='Group', detailed=True, data=df))
aov3_ss1 = df_aov3.anova(dv='Cholesterol',
between=['Sex', 'Drug'],
ss_type=1)
aov3_ss2 = df_aov3.anova(dv='Cholesterol',
between=['Sex', 'Drug'],
ss_type=2)
aov3_ss2_pg = pg.anova(dv='Cholesterol',
between=['Sex', 'Drug'],
data=df_aov3,
ss_type=2)
assert not aov3_ss1.equals(aov3_ss2)
assert aov3_ss2.round(3).equals(aov3_ss2_pg.round(3))
# Test the Welch ANOVA (Pandas)
aov = df.welch_anova(dv='Scores', between='Group')
assert aov.equals(pg.welch_anova(dv='Scores', between='Group',
data=df))
# Test the ANCOVA
aov = df_anc.ancova(dv='Scores', covar='Income',
between='Method').round(3)
assert (aov.equals(
pg.ancova(data=df_anc,
dv='Scores',
covar='Income',
between='Method').round(3)))
# Test the repeated measures ANOVA (Pandas)
aov = df.rm_anova(dv='Scores',
within='Time',
subject='Subject',
detailed=True)
assert (aov.equals(
pg.rm_anova(dv='Scores',
within='Time',
subject='Subject',
detailed=True,
data=df)))
# FDR-corrected post hocs with Hedges'g effect size
ttests = df.pairwise_tests(dv='Scores',
within='Time',
subject='Subject',
padjust='fdr_bh',
effsize='hedges')
assert (ttests.equals(
pg.pairwise_tests(dv='Scores',
within='Time',
subject='Subject',
padjust='fdr_bh',
effsize='hedges',
data=df)))
# Pairwise Tukey
tukey = df.pairwise_tukey(dv='Scores', between='Group')
assert tukey.equals(
pg.pairwise_tukey(data=df, dv='Scores', between='Group'))
# Test two-way mixed ANOVA
aov = df.mixed_anova(dv='Scores',
between='Group',
within='Time',
subject='Subject',
correction=False)
assert (aov.equals(
pg.mixed_anova(dv='Scores',
between='Group',
within='Time',
subject='Subject',
correction=False,
data=df)))
# Test parwise correlations
corrs = data.pairwise_corr(columns=['X', 'M', 'Y'], method='spearman')
corrs2 = pg.pairwise_corr(data=data,
columns=['X', 'M', 'Y'],
method='spearman')
assert corrs['r'].equals(corrs2['r'])
# Test partial correlation
corrs = data.partial_corr(x='X', y='Y', covar='M', method='spearman')
corrs2 = pg.partial_corr(x='X',
y='Y',
covar='M',
method='spearman',
data=data)
assert corrs['r'].equals(corrs2['r'])
# Test partial correlation matrix (compare with the ppcor package)
corrs = data.iloc[:, :5].pcorr().round(3)
np.testing.assert_array_equal(corrs.iloc[0, :].to_numpy(),
[1, 0.392, 0.06, -0.014, -0.149])
# Now compare against Pingouin's own partial_corr function
corrs = data[['X', 'Y', 'M']].pcorr()
corrs2 = data.partial_corr(x='X', y='Y', covar='M')
assert np.isclose(corrs.at['X', 'Y'], corrs2.at['pearson', 'r'])
# Test rcorr (correlation matrix with p-values)
# We compare against Pingouin pairwise_corr function
corrs = df_corr.rcorr(padjust='holm', decimals=4)
corrs2 = df_corr.pairwise_corr(padjust='holm').round(4)
assert corrs.at['Neuroticism', 'Agreeableness'] == '*'
assert (corrs.at['Agreeableness',
'Neuroticism'] == str(corrs2.at[2, 'r']))
corrs = df_corr.rcorr(padjust='holm', stars=False, decimals=4)
assert (corrs.at['Neuroticism',
'Agreeableness'] == str(corrs2.at[2,
'p-corr'].round(4)))
corrs = df_corr.rcorr(upper='n', decimals=5)
corrs2 = df_corr.pairwise_corr().round(5)
assert corrs.at['Extraversion', 'Openness'] == corrs2.at[4, 'n']
assert corrs.at['Openness', 'Extraversion'] == str(corrs2.at[4, 'r'])
# Method = spearman does not work with Python 3.5 on Travis?
# Instead it seems to return the Pearson correlation!
df_corr.rcorr(method='spearman')
df_corr.rcorr()
# Test mediation analysis
med = data.mediation_analysis(x='X', m='M', y='Y', seed=42, n_boot=500)
np.testing.assert_array_equal(med.loc[:, 'coef'].round(4).to_numpy(),
[0.5610, 0.6542, 0.3961, 0.0396, 0.3565])
ANOVApvals = np.empty((npatients))
allHstat = np.empty((npatients))
KWpvals = np.empty((npatients))
for p in range(npatients):
print(f'Loading summary delta data for Patient {p+1}')
summarydelta = sio.loadmat(f'data/Patient{p+1}_summarydelta')
stackeddelta = np.hstack((np.squeeze(summarydelta['state1delta']),
np.squeeze(summarydelta['state2delta'])))
whichstate = np.ones(stackeddelta.shape[0]) * 2
whichstate[0:(np.squeeze(summarydelta['state1delta']).size)] = 1
delta_df = pd.DataFrame({
'standarddelta': stackeddelta,
'brainstate': whichstate
})
aov = anova(dv='standarddelta',
between='brainstate',
data=delta_df,
detailed=True)
allFstat[p] = aov['F'][0]
ANOVApvals[p] = aov['p-unc'][0]
kw = kruskal(dv='standarddelta',
between='brainstate',
data=delta_df,
detailed=True)
allHstat[p] = kw['H'][0]
KWpvals[p] = kw['p-unc'][0]
nANOVAsig001 = np.sum(ANOVApvals < .001)
print(
f'There were {nANOVAsig001} significant differences by ANOVA (alpha = .001) of {npatients} patients between mean standardized delta across both brain-derived states'
)
nANOVAsig01 = np.sum(ANOVApvals < .01)
aov_table # %% [markdown] # The output from this command provides us with two things. First, it shows us the result of a t-test for each of the dummy variables, which basically tell us whether each of the conditions separately differs from placebo; it appears that Drug 1 does whereas Drug 2 does not. However, keep in mind that if we wanted to interpret these tests, we would need to correct the p-values to account for the fact that we have done multiple hypothesis tests; we will see an example of how to do this in the next chapter. # # Remember that the hypothesis that we started out wanting to test was whether there was any difference between any of the conditions; we refer to this as an *omnibus* hypothesis test, and it is the test that is provided by the F statistic. The F statistic basically tells us whether our model is better than a simple model that just includes an intercept. In this case we see that the F test is highly significant, consistent with our impression that there did seem to be differences between the groups (which in fact we know there were, because we created the data). # %% ols_model = ols(formula='BPsys~ group', data=df) ols_result = ols_model.fit() aov_table = sm.stats.anova_lm(ols_result) aov_table # %% import pingouin as pg pg.anova(data=df, dv='BPsys',between='group', effsize="np2") # %% pg.pairwise_tukey(data=df, dv='BPsys', between='group') # %% [markdown] # ## Learning objectives # # After reading this chapter, you should be able to: # # * Describe the rationale behind the sign test # * Describe how the t-test can be used to compare a single mean to a hypothesized value # * Compare the means for two paired or unpaired groups using a two-sample t-test # # # ## Appendix
