Skip to content

Commit

Permalink
uncommented target distr test
Browse files Browse the repository at this point in the history
  • Loading branch information
Nabil Fayak committed Aug 15, 2023
1 parent 2e4bc52 commit 9f2c0f0
Showing 1 changed file with 133 additions and 133 deletions.
266 changes: 133 additions & 133 deletions tests/data_checks_tests/test_target_distribution_data_check.py
View file
Original file line number Diff line number Diff line change
@@ -1,133 +1,133 @@
# import numpy as np
# import pandas as pd
# import pytest
# from scipy.stats import jarque_bera, lognorm, norm, shapiro

# from checkmates.data_checks import (
# DataCheckActionCode,
# DataCheckActionOption,
# DataCheckError,
# DataCheckMessageCode,
# DataCheckWarning,
# TargetDistributionDataCheck,
# )
# from checkmates.utils import infer_feature_types

# target_dist_check_name = TargetDistributionDataCheck.name


# def test_target_distribution_data_check_no_y(X_y_regression):
# X, y = X_y_regression
# y = None

# target_dist_check = TargetDistributionDataCheck()

# assert target_dist_check.validate(X, y) == [
# DataCheckError(
# message="Target is None",
# data_check_name=target_dist_check_name,
# message_code=DataCheckMessageCode.TARGET_IS_NONE,
# details={},
# ).to_dict(),
# ]


# @pytest.mark.parametrize("target_type", ["boolean", "categorical", "integer", "double"])
# def test_target_distribution_data_check_unsupported_target_type(target_type):
# X = pd.DataFrame(range(5))

# if target_type == "boolean":
# y = pd.Series([True, False] * 5)
# elif target_type == "categorical":
# y = pd.Series(["One", "Two", "Three", "Four", "Five"] * 2)
# elif target_type == "integer":
# y = [-1, -3, -5, 4, -2, 4, -4, 2, 1, 1]
# else:
# y = [9.2, 7.66, 4.93, 3.29, 4.06, -1.28, 4.95, 6.77, 9.07, 7.67]

# y = infer_feature_types(y)

# target_dist_check = TargetDistributionDataCheck()

# if target_type in ["integer", "double"]:
# assert target_dist_check.validate(X, y) == []
# else:
# assert target_dist_check.validate(X, y) == [
# DataCheckError(
# message=f"Target is unsupported {y.ww.logical_type.type_string} type. Valid Woodwork logical types include: integer, double, age, age_fractional",
# data_check_name=target_dist_check_name,
# message_code=DataCheckMessageCode.TARGET_UNSUPPORTED_TYPE,
# details={"unsupported_type": y.ww.logical_type.type_string},
# ).to_dict(),
# ]


# @pytest.mark.parametrize("data_type", ["positive", "mixed", "negative"])
# @pytest.mark.parametrize("distribution", ["normal", "lognormal", "very_lognormal"])
# @pytest.mark.parametrize(
# "size,name,statistic",
# [(10000, "jarque_bera", jarque_bera), (5000, "shapiro", shapiro)],
# )
# def test_target_distribution_data_check_warning_action(
# size,
# name,
# statistic,
# distribution,
# data_type,
# X_y_regression,
# ):
# X, y = X_y_regression
# # set this to avoid flaky tests. This is primarily because when we have smaller samples,
# # once we remove values outside 3 st.devs, the distribution can begin to look more normal
# random_state = 2
# target_dist_check = TargetDistributionDataCheck()

# if distribution == "normal":
# y = norm.rvs(loc=3, size=size, random_state=random_state)
# elif distribution == "lognormal":
# y = lognorm.rvs(0.4, size=size, random_state=random_state)
# else:
# # Will have a p-value of 0 thereby rejecting the null hypothesis even after log transforming
# # This is essentially just checking the = of the statistic's "log.pvalue >= og.pvalue"
# y = lognorm.rvs(s=1, loc=1, scale=1, size=size, random_state=random_state)

# y = np.round(y, 6)

# if data_type == "negative":
# y = -np.abs(y)
# elif data_type == "mixed":
# y = y - 1.2

# if distribution == "normal":
# assert target_dist_check.validate(X, y) == []
# else:
# target_dist_ = target_dist_check.validate(X, y)

# if any(y <= 0):
# y = y + abs(y.min()) + 1
# y = y[y < (y.mean() + 3 * round(y.std(), 3))]
# test_og = statistic(y)

# details = {
# "normalization_method": name,
# "statistic": round(test_og.statistic, 1),
# "p-value": round(test_og.pvalue, 3),
# }
# assert target_dist_ == [
# DataCheckWarning(
# message="Target may have a lognormal distribution.",
# data_check_name=target_dist_check_name,
# message_code=DataCheckMessageCode.TARGET_LOGNORMAL_DISTRIBUTION,
# details=details,
# action_options=[
# DataCheckActionOption(
# DataCheckActionCode.TRANSFORM_TARGET,
# data_check_name=target_dist_check_name,
# metadata={
# "is_target": True,
# "transformation_strategy": "lognormal",
# },
# ),
# ],
# ).to_dict(),
# ]
import numpy as np
import pandas as pd
import pytest
from scipy.stats import jarque_bera, lognorm, norm, shapiro

from checkmates.data_checks import (
DataCheckActionCode,
DataCheckActionOption,
DataCheckError,
DataCheckMessageCode,
DataCheckWarning,
TargetDistributionDataCheck,
)
from checkmates.utils import infer_feature_types

target_dist_check_name = TargetDistributionDataCheck.name


def test_target_distribution_data_check_no_y(X_y_regression):
X, y = X_y_regression
y = None

target_dist_check = TargetDistributionDataCheck()

assert target_dist_check.validate(X, y) == [
DataCheckError(
message="Target is None",
data_check_name=target_dist_check_name,
message_code=DataCheckMessageCode.TARGET_IS_NONE,
details={},
).to_dict(),
]


@pytest.mark.parametrize("target_type", ["boolean", "categorical", "integer", "double"])
def test_target_distribution_data_check_unsupported_target_type(target_type):
X = pd.DataFrame(range(5))

if target_type == "boolean":
y = pd.Series([True, False] * 5)
elif target_type == "categorical":
y = pd.Series(["One", "Two", "Three", "Four", "Five"] * 2)
elif target_type == "integer":
y = [-1, -3, -5, 4, -2, 4, -4, 2, 1, 1]
else:
y = [9.2, 7.66, 4.93, 3.29, 4.06, -1.28, 4.95, 6.77, 9.07, 7.67]

y = infer_feature_types(y)

target_dist_check = TargetDistributionDataCheck()

if target_type in ["integer", "double"]:
assert target_dist_check.validate(X, y) == []
else:
assert target_dist_check.validate(X, y) == [
DataCheckError(
message=f"Target is unsupported {y.ww.logical_type.type_string} type. Valid Woodwork logical types include: integer, double, age, age_fractional",
data_check_name=target_dist_check_name,
message_code=DataCheckMessageCode.TARGET_UNSUPPORTED_TYPE,
details={"unsupported_type": y.ww.logical_type.type_string},
).to_dict(),
]


@pytest.mark.parametrize("data_type", ["positive", "mixed", "negative"])
@pytest.mark.parametrize("distribution", ["normal", "lognormal", "very_lognormal"])
@pytest.mark.parametrize(
"size,name,statistic",
[(10000, "jarque_bera", jarque_bera), (5000, "shapiro", shapiro)],
)
def test_target_distribution_data_check_warning_action(
size,
name,
statistic,
distribution,
data_type,
X_y_regression,
):
X, y = X_y_regression
# set this to avoid flaky tests. This is primarily because when we have smaller samples,
# once we remove values outside 3 st.devs, the distribution can begin to look more normal
random_state = 2
target_dist_check = TargetDistributionDataCheck()

if distribution == "normal":
y = norm.rvs(loc=3, size=size, random_state=random_state)
elif distribution == "lognormal":
y = lognorm.rvs(0.4, size=size, random_state=random_state)
else:
# Will have a p-value of 0 thereby rejecting the null hypothesis even after log transforming
# This is essentially just checking the = of the statistic's "log.pvalue >= og.pvalue"
y = lognorm.rvs(s=1, loc=1, scale=1, size=size, random_state=random_state)

y = np.round(y, 6)

if data_type == "negative":
y = -np.abs(y)
elif data_type == "mixed":
y = y - 1.2

if distribution == "normal":
assert target_dist_check.validate(X, y) == []
else:
target_dist_ = target_dist_check.validate(X, y)

if any(y <= 0):
y = y + abs(y.min()) + 1
y = y[y < (y.mean() + 3 * round(y.std(), 3))]
test_og = statistic(y)

details = {
"normalization_method": name,
"statistic": round(test_og.statistic, 1),
"p-value": round(test_og.pvalue, 3),
}
assert target_dist_ == [
DataCheckWarning(
message="Target may have a lognormal distribution.",
data_check_name=target_dist_check_name,
message_code=DataCheckMessageCode.TARGET_LOGNORMAL_DISTRIBUTION,
details=details,
action_options=[
DataCheckActionOption(
DataCheckActionCode.TRANSFORM_TARGET,
data_check_name=target_dist_check_name,
metadata={
"is_target": True,
"transformation_strategy": "lognormal",
},
),
],
).to_dict(),
]

0 comments on commit 9f2c0f0

Please sign in to comment.