def test_RSI_wrong_type(self):
"""
Negative test
data: array (unsupported type)
Checks that the random_sample_imputation raises a TypeError if the
data is passed as an array.
"""
# 1. Arrange
data = [2, 4, np.nan, 1]
# 2. Act & 3. Assert
with self.assertRaises(TypeError):
random_sample_imputation(data)
def test_RSI_series_inplace(self):
"""
Positive test
data: Correct Series (example series)
Checks that random_sample_imputation removes 3 NA values from the
series.
"""
# 1. Arrange
ser = generate_example_series()
# 2. Act
random_sample_imputation(ser, inplace=True)
# 3. Assert
self.assertEqual(ser.isna().sum(), 0)
def test_RSI_df_inplace(self):
"""
Positive test
data: Correct dataframe (divcols)
Checks that random_sample_imputation removes 8 NA values from the
dataframe.
"""
# 1. Arrange
df = generate_example_df_divcols()
# 2. Act
random_sample_imputation(df, inplace=True)
# 3. Assert
self.assertEqual(df.isna().sum().sum(), 10)
def test_NOCB_df_inplace_wrong_column(self):
"""
Negative test
data: Correct dataframe (divcols)
columns: ['f', 'g', 'z'] ('z' doesn't exist in the data)
Checks that the random_sample_imputation raises a ValueError if one of
the specified columns doesn't exist in the data.
"""
# 1. Arrange
df = generate_example_df_divcols()
# 2. Act & 3. Assert
with self.assertRaises(ValueError):
random_sample_imputation(df, columns=['f', 'g', 'z'], inplace=True)
def test_RSI_col_for_series(self):
"""
Negative test
data: Correct series (example_series)
columns: ['a'] (series can't have columns)
Checks that the function raises a ValueError if a column is passed
for a series.
"""
# 1. Arrange
ser = generate_example_series()
# 2. Act & 3. Assert
with self.assertRaises(ValueError):
random_sample_imputation(ser, columns=['a'])
def test_RSI_df_inplace_columns(self):
"""
Positive test
data: Correct dataframe (divcols)
columns: ['f', 'g']
Checks that random_sample_imputation removes 4 NA values from the
specified columns.
"""
# 1. Arrange
df = generate_example_df_divcols()
# 2. Act
random_sample_imputation(df, columns=['f', 'g'], inplace=True)
# 3. Assert
self.assertEqual(df.isna().sum().sum(), 14)
def mice_one_imputation(data):
"""Auxiliary function that performs one MICE imputation, choosing the
order in which the columns are imputed at random.
:param data: The data on which to perform the imputation.
:type data: pandas.DataFrame
:return: The dataframe with one MICE imputation performed.
:rtype: pandas.DataFrame
"""
# This auxiliary function always returns a copy:
res = data.copy()
# Save the mask of missing values:
na_mask = pd.isna(data)
# Compute the list of columns with missing values
columns_with_na = []
for column in data.columns:
if data[column].isna().any():
columns_with_na.append(column)
# Shuffle the list of columns to impute:
shuffle(columns_with_na)
# Impute with mean substitution:
for column in columns_with_na:
if is_numeric_dtype(data[column]):
mean_substitution(res, columns=[column], inplace=True)
else:
random_sample_imputation(res, columns=[column], inplace=True)
# Compute which columns are numeric in order to use them as predictors:
numerics = [col for col in data.columns if is_numeric_dtype(data[col])]
# Impute each column:
for column in columns_with_na:
if is_numeric_dtype(data[column]):
res.loc[na_mask[column], column] = np.nan
linear_regression(res, column, predictors=numerics, inplace=True)
else:
res.loc[na_mask[column], column] = None
logistic_regression(res, column, inplace=True)
return res
def test_RSI_series_returning(self):
"""
Positive test
data: Correct Series (example series)
Checks that the original series remains unmodified and that the
returned series contains no NA values, 3 less than the original.
"""
# 1. Arrange
ser = generate_example_series()
# 2. Act
ser2 = random_sample_imputation(ser)
# 3. Assert
self.assertEqual(ser.isna().sum(), 3)
self.assertEqual(ser2.isna().sum(), 0)
def test_RSI_df_returning(self):
"""
Positive test
data: Correct dataframe (divcols)
Checks that the original dataframe remains unmodified and that the
returned dataframe contains 10 NA values, 8 less than the original.
"""
# 1. Arrange
df = generate_example_df_divcols()
# 2. Act
df2 = random_sample_imputation(df)
# 3. Assert
self.assertEqual(df.isna().sum().sum(), 18)
self.assertEqual(df2.isna().sum().sum(), 10)