def test_one_sample_hypothesis():
"""Check one sample ttest."""
schema = DataFrameSchema({
"height_in_feet": Column(
Float, [
Hypothesis.one_sample_ttest(
popmean=5,
relationship="greater_than",
alpha=0.1),
]
),
})
subset_schema = DataFrameSchema({
"group": Column(String),
"height_in_feet": Column(
Float, [
Hypothesis.one_sample_ttest(
sample="A",
groupby="group",
popmean=5,
relationship="greater_than",
alpha=0.1),
]
),
})
df = (
pd.DataFrame({
"height_in_feet": [8.1, 7, 6.5, 6.7, 5.1],
"group": ["A", "A", "B", "B", "A"],
})
)
schema.validate(df)
subset_schema.validate(df)
def test_two_sample_ttest_hypothesis_relationships():
"""Check allowable relationships in two-sample ttest."""
for relationship in Hypothesis.RELATIONSHIPS:
schema = DataFrameSchema({
"height_in_feet": Column(Float, [
Hypothesis.two_sample_ttest(
sample1="M",
sample2="F",
groupby="sex",
relationship=relationship,
alpha=0.5),
]),
"sex": Column(String)
})
assert isinstance(schema, DataFrameSchema)
for relationship in ["foo", "bar", 1, 2, 3, None]:
with pytest.raises(errors.SchemaInitError):
DataFrameSchema({
"height_in_feet": Column(Float, [
Hypothesis.two_sample_ttest(
sample1="M",
sample2="F",
groupby="sex",
relationship=relationship,
alpha=0.5),
]),
"sex": Column(String)
})
def test_dataframe_hypothesis_checks():
"""
Test that two specific implementations of a Hypothesis work as expected
and that using a Column that wasn't defined will error.
"""
df = pd.DataFrame({
"col1": range(100, 201),
"col2": range(0, 101),
})
hypothesis_check_schema = DataFrameSchema(
columns={
"col1": Column(Int),
"col2": Column(Int),
},
checks=[
# two-sample test
Hypothesis(
test=stats.ttest_ind,
samples=["col1", "col2"],
relationship=lambda stat, pvalue, alpha=0.01:
(stat > 0 and pvalue / 2 < alpha),
relationship_kwargs={"alpha": 0.5},
),
# one-sample test
Hypothesis(
test=stats.ttest_1samp,
samples=["col1"],
relationship=lambda stat, pvalue, alpha=0.01:
(stat > 0 and pvalue / 2 < alpha),
test_kwargs={"popmean": 50},
relationship_kwargs={"alpha": 0.01},
),
],
)
hypothesis_check_schema.validate(df)
# raise error when using groupby for a column that doesn't exist
hypothesis_check_schema_groupby = DataFrameSchema(
columns={
"col1": Column(Int),
"col2": Column(Int),
},
checks=[
# two-sample test
Hypothesis(
test=stats.ttest_ind,
samples=["col1", "col2"],
groupby="col3",
relationship=lambda stat, pvalue, alpha=0.01:
(stat > 0 and pvalue / 2 < alpha),
relationship_kwargs={"alpha": 0.5},
),
],
)
with pytest.raises(errors.SchemaDefinitionError):
hypothesis_check_schema_groupby.validate(df)
def test_dataframe_hypothesis_checks():
df = pd.DataFrame({
"col1": range(100, 201),
"col2": range(0, 101),
})
hypothesis_check_schema = DataFrameSchema(
columns={
"col1": Column(Int),
"col2": Column(Int),
},
checks=[
# two-sample test
Hypothesis(
test=stats.ttest_ind,
samples=["col1", "col2"],
relationship=lambda stat, pvalue, alpha=0.01: (
stat > 0 and pvalue / 2 < alpha
),
relationship_kwargs={"alpha": 0.5},
),
# one-sample test
Hypothesis(
test=stats.ttest_1samp,
samples=["col1"],
relationship=lambda stat, pvalue, alpha=0.01: (
stat > 0 and pvalue / 2 < alpha
),
test_kwargs={"popmean": 50},
relationship_kwargs={"alpha": 0.01},
),
]
)
hypothesis_check_schema.validate(df)
# raise error when using groupby
hypothesis_check_schema_groupby = DataFrameSchema(
columns={
"col1": Column(Int),
"col2": Column(Int),
},
checks=[
# two-sample test
Hypothesis(
test=stats.ttest_ind,
samples=["col1", "col2"],
groupby="col3",
relationship=lambda stat, pvalue, alpha=0.01: (
stat > 0 and pvalue / 2 < alpha
),
relationship_kwargs={"alpha": 0.5},
),
]
)
with pytest.raises(errors.SchemaDefinitionError):
hypothesis_check_schema_groupby.validate(df)
def hypothesis_accident_probability(feature, increases=True):
relationship = "greater_than" if increases else "less_than"
return {
feature: Column(checks=Check.isin([1, 0])),
f"{feature}_shap": Column(
pa.Float,
checks=Hypothesis.two_sample_ttest(
sample1=1,
sample2=0,
groupby=feature,
relationship=relationship,
alpha=0.01,
)
),
}
def test_hypothesis():
"""Tests the different API calls of Hypothesis."""
# Example df for tests:
df = pd.DataFrame({
"height_in_feet": [6.5, 7, 6.1, 5.1, 4],
"sex": ["M", "M", "F", "F", "F"],
})
# Initialise the different ways of calling a test:
schema_pass_ttest_on_alpha_val_1 = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis.two_sample_ttest(
sample1="M",
sample2="F",
groupby="sex",
relationship="greater_than",
alpha=0.5,
),
],
),
"sex":
Column(String),
})
schema_pass_ttest_on_alpha_val_2 = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis(
test=stats.ttest_ind,
samples=["M", "F"],
groupby="sex",
relationship="greater_than",
relationship_kwargs={"alpha": 0.5},
),
],
),
"sex":
Column(String),
})
schema_pass_ttest_on_alpha_val_3 = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis.two_sample_ttest(
sample1="M",
sample2="F",
groupby="sex",
relationship="greater_than",
alpha=0.5,
),
],
),
"sex":
Column(String),
})
schema_pass_ttest_on_custom_relationship = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis(
test=stats.ttest_ind,
samples=["M", "F"],
groupby="sex",
relationship=lambda stat, pvalue, alpha=0.01:
(stat > 0 and pvalue / 2 < alpha),
relationship_kwargs={"alpha": 0.5},
)
],
),
"sex":
Column(String),
})
# Check the 3 happy paths are successful:
schema_pass_ttest_on_alpha_val_1.validate(df)
schema_pass_ttest_on_alpha_val_2.validate(df)
schema_pass_ttest_on_alpha_val_3.validate(df)
schema_pass_ttest_on_custom_relationship.validate(df)
schema_fail_ttest_on_alpha_val_1 = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis.two_sample_ttest(
sample1="M",
sample2="F",
groupby="sex",
relationship="greater_than",
alpha=0.05,
),
],
),
"sex":
Column(String),
})
schema_fail_ttest_on_alpha_val_2 = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis(
test=stats.ttest_ind,
samples=["M", "F"],
groupby="sex",
relationship="greater_than",
relationship_kwargs={"alpha": 0.05},
),
],
),
"sex":
Column(String),
})
schema_fail_ttest_on_alpha_val_3 = DataFrameSchema({
"height_in_feet":
Column(
Float,
[
Hypothesis.two_sample_ttest(
sample1="M",
sample2="F",
groupby="sex",
relationship="greater_than",
alpha=0.05,
),
],
),
"sex":
Column(String),
})
with pytest.raises(errors.SchemaError):
schema_fail_ttest_on_alpha_val_1.validate(df)
with pytest.raises(errors.SchemaError):
schema_fail_ttest_on_alpha_val_2.validate(df)
with pytest.raises(errors.SchemaError):
schema_fail_ttest_on_alpha_val_3.validate(df)
percent_accidental = fatal_encounters_clean.disposition_accidental.mean()
display(Markdown(f"{percent_accidental * 100:0.02f}%"))
# %% [markdown] slideshow={"slide_type": "subslide"}
# **Hypothesis**: "the `disposition_accidental` target has a
# class balance of ~2.75%"
# %%
from pandera import Hypothesis
# use the Column object as a stand-alone schema object
target_schema = Column(
pa.Bool,
name="disposition_accidental",
checks=Hypothesis.one_sample_ttest(
popmean=0.0275, relationship="equal", alpha=0.01
)
)
target_schema(fatal_encounters_clean);
# %% [markdown] slideshow={"slide_type": "skip"}
# What's the age distribution in the dataset?
# %% tags=["hide_input"] slideshow={"slide_type": "skip"}
fatal_encounters_clean.age.plot.hist(figsize=(8, 5)).set_xlabel("age");
# %% [markdown] slideshow={"slide_type": "skip"}
# **Hypothesis** check: "the `age` column is right-skewed"
# %% slideshow={"slide_type": "skip"}
