P values for sklearn logistic regression

p_values_for_logreg.py

from sklearn import linear_model
import numpy as np
import scipy.stats as stat

class LogisticReg:
    """
    Wrapper Class for Logistic Regression which has the usual sklearn instance 
    in an attribute self.model, and pvalues, z scores and estimated 
    errors for each coefficient in 
    
    self.z_scores
    self.p_values
    self.sigma_estimates
    
    as well as the negative hessian of the log Likelihood (Fisher information)
    
    self.F_ij
    """
    
    def __init__(self,*args,**kwargs):#,**kwargs):
        self.model = linear_model.LogisticRegression(*args,**kwargs)#,**args)

    def fit(self,X,y):
        self.model.fit(X,y)
        #### Get p-values for the fitted model ####
        denom = (2.0*(1.0+np.cosh(self.model.decision_function(X))))
        denom = np.tile(denom,(X.shape[1],1)).T
        F_ij = np.dot((X/denom).T,X) ## Fisher Information Matrix
        Cramer_Rao = np.linalg.inv(F_ij) ## Inverse Information Matrix
        sigma_estimates = np.sqrt(np.diagonal(Cramer_Rao))
        z_scores = self.model.coef_[0]/sigma_estimates # z-score for eaach model coefficient
        p_values = [stat.norm.sf(abs(x))*2 for x in z_scores] ### two tailed test for p-values
        
        self.z_scores = z_scores
        self.p_values = p_values
        self.sigma_estimates = sigma_estimates
        self.F_ij = F_ij
        

Hi @biohousten , your code was very helpful, but I am getting an error claiming 'dep_acute' is not defined, can you help me understand where this variable comes from? Cheers

It boggles my mind that it is so freaking complicated to get a p-value in sklearn. My God, what is the freaking point of doing logistic regression if you can't assess whether your p-values are significant or not? Why is this not a basic function argument?