Resampling

from functools import partial
from sklearn.model_selection import \
     (cross_validate,
      KFold,
      ShuffleSplit)
from sklearn.base import clone
from ISLP.models import sklearn_sm

def evalMSE(terms,
            response,
            train,
            test):

   mm = MS(terms)
   X_train = mm.fit_transform(train)
   y_train = train[response]

   X_test = mm.transform(test)
   y_test = test[response]

   results = sm.OLS(y_train, X_train).fit()
   test_pred = results.predict(X_test)

   return np.mean((y_test - test_pred)**2)

Cross-Validation

• use ISLP wrapper sklearn_sm() because sklearn and statsmodels don't play nice

Bootstrap

def boot_SE(func,
            D,
            n=None,
            B=1000,
            seed=0):
    rng = np.random.default_rng(seed)
    first_, second_ = 0, 0
    n = n or D.shape[0]
    for _ in range(B):
        idx = rng.choice(D.index,
                         n,
                         replace=True)
        value = func(D, idx)
        first_ += value
        second_ += value**2
    return np.sqrt(second_ / B - (first_ / B)**2)

def boot_OLS(model_matrix, response, D, idx):
    D_ = D.loc[idx]
    Y_ = D_[response]
    X_ = clone(model_matrix).fit_transform(D_)
    return sm.OLS(Y_, X_).fit().params

• partial() takes a function as an argument, and freezes some of its arguments, starting from the left.