def rm_one_way_anova(dataset: Dataset, design, combined_data: CombinedData):
data = dataset.data
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(ys) == 1)
y = ys[0]
between_subjs = []
within_subjs = []
for x in xs:
if "between subjects" in design and design[
"between subjects"] == x.metadata[name]:
between_subjs.append(x.metadata[name])
if "within subjects" in design and design[
"within subjects"] == x.metadata[name]:
within_subjs.append(x.metadata[name])
# import pdb; pdb.set_trace()
id = dataset.pid_col_name
aovrm2way = AnovaRM(data,
depvar=y.metadata[name],
subject=id,
within=within_subjs)
# aovrm2way = AnovaRM(data, depvar=y.metadata[name], subject=dataset.pid_col_name, within=within_subjs, between=between_subjs) # apparently not implemented in statsmodels
# import pdb; pdb.set_trace()
res2way = aovrm2way.fit()
def bootstrap(dataset: Dataset, combined_data: CombinedData):
calculations = {}
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
for y in ys:
# for now
assert (len(ys) == 1)
# Main effects
for x in xs:
cat = [k for k, v in x.metadata[categories].items()]
for c in cat:
cat_data = dataset.select(
y.metadata[name], where=[f"{x.metadata[name]} == '{c}'"])
stat = bs.bootstrap(cat_data.to_numpy(),
stat_func=bs_stats.median)
calculations[c] = stat
# import pdb; pdb.set_trace()
# store all the medians & confidence intervals
# return all the medians & CIs
# data.append(cat_data)
return calculations
def factorial_ANOVA(dataset: Dataset, combined_data: CombinedData):
# Construct formula
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(ys) == 1)
y = ys[0]
formula = f"{y.metadata[name]} ~ "
for i in range(len(xs)):
x = xs[i]
formula += f"C({x.metadata[name]})"
if i < len(xs) - 1:
formula += " + "
# Add the interactions
interactions = []
for i in range(len(xs)):
x_i = xs[i]
inter = f"C({x_i.metadata[name]})"
for j in range(len(xs)):
if i != j:
x_j = xs[j]
inter += " * " + f"C({x_j.metadata[name]})"
interactions.append(inter)
if _is_interaction_unique(interactions, inter):
formula += " + " + inter
ols_formula = ols(formula, data=dataset.data)
model = ols_formula.fit()
return sm.stats.anova_lm(model, type=2)
def chi_square(dataset: Dataset, combined_data: CombinedData):
# Compute the contingency table
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
if len(xs) == 1:
if len(ys) == 1:
x = xs[0]
y = ys[0]
# Get the count for each category
x_cat = [k for k, v in x.metadata[categories].items()]
y_cat = [k for k, v in y.metadata[categories].items()]
contingency_table = []
contingency_table_key = [
] # labels for the order in which data is stored in data array (define above)
for xc in x_cat:
table_row = []
table_row_key = []
for yc in y_cat:
data = dataset.select(y.metadata[name],
where=[
f"{x.metadata[name]} == '{xc}'",
f"{y.metadata[name]} == '{yc}'"
])
table_row.append(len(data))
x_y_key = str(
x.metadata[name]) + ':' + str(xc) + ' by ' + str(
y.metadata[name]) + ':' + str(yc)
table_row_key.append(x_y_key)
assert (len(table_row_key) == len(table_row))
assert (len(table_row) == len(y_cat))
contingency_table.append(table_row)
contingency_table_key.append(table_row_key)
else:
raise ValueError(
f"Currently, chi square requires/only supports 1 explained variable, instead received: {len(ys)} -- {ys}"
)
else:
raise ValueError(
f"Currently, chi square requires/only supports 1 explanatory variable, instead received: {len(xs)} -- {xs}"
)
# chi2, p, dof, ex = chi2_contingency(obs, correction=False)
chi2, p, dof, ex = stats.chi2_contingency(contingency_table,
correction=False)
return ChisquareResult(chi2, p, dof, ex)
def f_test(dataset: Dataset, combined_data: CombinedData):
# Construct formula
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(xs) == 1)
assert (len(ys) == 1)
x = xs[0]
y = ys[0]
formula = ols(f"{y.metadata[name]} ~ C({x.metadata[name]})",
data=dataset.data)
model = formula.fit()
return sm.stats.anova_lm(model, type=2)
def wilcoxon_signed_rank(dataset: Dataset, combined_data: CombinedData):
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
x = xs[0]
y = ys[0]
cat = [k for k, v in x.metadata[categories].items()]
data = []
for c in cat:
cat_data = dataset.select(y.metadata[name],
where=[f"{x.metadata[name]} == '{c}'"])
data.append(cat_data)
return stats.wilcoxon(data[0], data[1])
def paired_students_t(dataset, combined_data: CombinedData):
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
x = xs[0]
y = ys[0]
cat = [k for k, v in x.metadata[categories].items()]
data = []
for c in cat:
cat_data = dataset.select(y.metadata[name],
where=[f"{x.metadata[name]} == '{c}'"])
data.append(cat_data)
return stats.ttest_rel(data[0], data[1])
def friedman(dataset: Dataset, combined_data: CombinedData):
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(ys) == 1)
y = ys[0]
data = []
for x in xs:
cat = [k for k, v in x.metadata[categories].items()]
for c in cat:
cat_data = dataset.select(y.metadata[name],
where=[f"{x.metadata[name]} == '{c}'"])
data.append(cat_data)
return stats.friedmanchisquare(*data)
def greater_than_5_frequency(dataset: Dataset, var_data: CombinedData, alpha):
xs = var_data.get_explanatory_variables()
ys = var_data.get_explained_variables()
if len(xs) == 1:
if len(ys) == 1:
x = xs[0]
y = ys[0]
if x.is_categorical() and y.is_categorical():
# Get the count for each category
x_cat = [k for k,v in x.metadata[categories].items()]
y_cat = [k for k,v in y.metadata[categories].items()]
for xc in x_cat:
for yc in y_cat:
data = dataset.select(y.metadata[name], where=[f"{x.metadata[name]} == '{xc}'", f"{y.metadata[name]} == '{yc}'"])
# Check that the count is at least five for each of the (x,y) group pairs
if (len(data) < 5):
return False
return True
else:
return False
else:
raise ValueError(f"Currently, chi square requires/only supports 1 explained variable, instead received: {len(ys)} -- {ys}")
else:
x0 = xs[0]
x1 = xs[1]
if x0.is_categorical() and x1.is_categorical():
# Get the count for each category
x0_cat = [k for k,v in x0.metadata[categories].items()]
x1_cat = [k for k,v in x1.metadata[categories].items()]
for x0c in x0_cat:
for x1c in x1_cat:
data = dataset.select(x1.metadata[name], where=[f"{x.metadata[name]} == '{xc}'", f"{x1.metadata[name]} == '{x1c}'"])
# Check that the count is at least five for each of the (x,x1) group pairs
if (len(data) < 5):
return False
return True
else:
return False
def pointbiserial(dataset: Dataset, combined_data: CombinedData):
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(xs) == 1)
assert (len(ys) == 1)
x = xs[0]
y = ys[0]
cat = [k for k, v in x.metadata[categories].items()]
data = []
for c in cat:
cat_data = dataset.select(y.metadata[name],
where=[f"{x.metadata[name]} == '{c}'"])
data.append(cat_data)
return stats.pointbiserialr(data[0], data[1])
def kruskall_wallis(dataset: Dataset, combined_data: CombinedData):
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(ys) == 1)
y = ys[0]
data = []
for x in xs:
if x.metadata[categories] is None:
import pdb
pdb.set_trace()
cat = [k for k, v in x.metadata[categories].items()]
for c in cat:
cat_data = dataset.select(y.metadata[name],
where=[f"{x.metadata[name]} == '{c}'"])
data.append(cat_data)
return stats.kruskal(*data)
def fishers_exact(dataset: Dataset, combined_data: CombinedData):
assert (len(combined_data.vars) == 2)
# Compute the contingency table
xs = combined_data.get_explanatory_variables()
ys = combined_data.get_explained_variables()
assert (len(xs) == 1)
assert (len(ys) == 1)
x = xs[0]
y = ys[0]
# Get the count for each category
x_cat = [k for k, v in x.metadata[categories].items()]
y_cat = [k for k, v in y.metadata[categories].items()]
contingency_table = []
contingency_table_key = [
] # labels for the order in which data is stored in data array (define above)
for xc in x_cat:
table_row = []
table_row_key = []
for yc in y_cat:
data = dataset.select(y.metadata[name],
where=[
f"{x.metadata[name]} == '{xc}'",
f"{y.metadata[name]} == '{yc}'"
])
table_row.append(len(data))
x_y_key = str(x.metadata[name]) + ':' + str(xc) + ' by ' + str(
y.metadata[name]) + ':' + str(yc)
table_row_key.append(x_y_key)
assert (len(table_row_key) == len(table_row))
assert (len(table_row) == len(y_cat))
contingency_table.append(table_row)
contingency_table_key.append(table_row_key)
odds_ratio, p_value = stats.fisher_exact(contingency_table,
alternative='two-sided')
return FishersResult(odds_ratio, p_value)
def has_equal_variance(dataset: Dataset, var_data: CombinedData, alpha):
xs = var_data.get_explanatory_variables()
ys = var_data.get_explained_variables()
cat_xs = []
cont_ys = []
grouped_data = []
for x in xs:
if x.is_categorical():
cat_xs.append(x)
for y in ys:
if y.is_continuous():
cont_ys.append(y)
eq_var = (None, None)
if cat_xs and cont_ys:
for y in ys:
for x in xs:
cat = [k for k,v in x.metadata[categories].items()]
for c in cat:
data = dataset.select(y.metadata[name], where=[f"{x.metadata[name]} == '{c}'"])
grouped_data.append(data)
if isinstance(var_data, BivariateData):
# Equal variance
eq_var = compute_eq_variance(grouped_data)
# elif isinstance(var_data, MultivariateData):
# var_data.properties[eq_variance + '::' + x.metadata[name] + ':' + y.metadata[name]] = compute_eq_variance(grouped_data)
else:
raise ValueError(f"var_data_data object is neither BivariateData nor MultivariateData: {type(var_data)}")
if eq_var[0] is None and eq_var[1] is None:
import pdb; pdb.set_trace()
# raise Exception("did not compute variance, this is a bug")
return False
return (eq_var[1] > alpha)
def has_one_x(dataset: Dataset, var_data: CombinedData, alpha):
xs = var_data.get_explanatory_variables()
return len(xs) == 1
