stasSajin

def estimate_rf(df):
  X = df.income.values.reshape(-1,1)
  y = df.minutes.values
  rf_reg = RandomForestRegressor(100, n_jobs=8, oob_score=True)
  rf_reg.fit(X,y)
  # store residuals
  y_new = df.minutes - reg.oob_prediction_
  return y_new[df[group == 1]].mean() - y_new[df[group == 0]].mean()

stasSajin

def estimate_cuped(df):
  # estimate the theta that minimizes the control variate
  theta = df.cov().loc["income", "minutes"] / df.cov().loc["income", "income"]
  y_new = df.minutes - theta * df.income
  return y_new[df[group == 1]].mean() - y_new[df[group == 0]].mean()

stasSajin

def generate_data(group_size, effect_size = 0):
    group1 = np.array([1] * group_size)
    group2 = np.array([0] * group_size)
    minutes = np.maximum(0, np.random.normal(25, 20, size = 2*group_size).astype(int))
    income = (np.random.normal(10000, 2000, size = 2*group_size) * minutes).astype(int)
    assets = (np.random.normal(100000, 20000, size = 2*group_size) * minutes).astype(int)
    group_assignment = np.concatenate([group1, group2])
    np.random.shuffle(group_assignment)
    data = pd.DataFrame({'group': group_assignment,
        'minutes': minutes,

stasSajin

from typing import Union, Dict
from enum import Enum
from pydantic.dataclasses import dataclass
from pydantic import ValidationError, validator
from pydantic import confloat, conint

class ProcedureType(Enum):
    cancer = "cancer"
    flu = "flu"
    

stasSajin

from hypothesis import strategies as st
from hypothesis import given
import pytest
@given(data=st.from_type(UserAssessment))
def test_apply_treatment(data):
    result = apply_treatment(data)
    # specify asserts in here

stasSajin

from typing import Optional, Union
from enum import Enum
from dataclasses import dataclass
from pydantic import ValidationError


class ProcedureType(Enum):
    cancer = "cancer"
    flu = "flu"

stasSajin

bias_var = np.array('gender_male')
bias_idx = np.argwhere(np.isin(np.array(feature_names), bias_var))[0]
interactions[:, bias_idx, :] = 0
interactions[:, :, bias_idx] = 0
y_hat_no_bias = np.exp(interactions.sum(axis = 1).sum(axis = 1)) - 1

stasSajin

feature_names = dtest.feature_names
interactions = xgb_model.predict(dtest, pred_interactions=True)
pd.DataFrame(interactions[0],
             index = feature_names + ['intercept'],
             columns= feature_names + ['intercept'])

stasSajin

import pandas as pd
import numpy as np
import xgboost as xgb
from sklearn.model_selection import train_test_split

data = pd.read_feather('data.feather')
data = pd.get_dummies(data, drop_first=True)
X_train, X_test, y_train, y_test = train_test_split(data.drop('salary', axis = 1),
                                                    np.log1p(data.salary.values),
                                                    test_size =.30,

stasSajin

Keybase proof

I hereby claim:

• I am stassajin on github.
• I am stassajin (https://keybase.io/stassajin) on keybase.
• I have a public key ASCgFOn7OF-Xnk4rKYYLqm6FUhEGDxCmg9ba3Vd6ecGWTAo

To claim this, I am signing this object: