0

我正在尝试使用最大似然方法来估计多重线性概率模型（killer_apps 是有限因变量）。因此，在这段代码中，我试图估计回归量和各自的标准误差。谁能告诉我为什么我会收到这个错误，一般来说，我做错了什么？我试图从用于简单线性概率模型的示例中扩展代码

def probitlik(beta,y,x1,x2,x3):
      beta1 = beta[0]
      beta2 = beta[1]
      beta3 = beta[2]
      beta4 = beta[3]
      prob = stats.norm.cdf(beta1 + beta2*x1 + beta3*x2 + beta4*x3)
      vlog=y*np.log(prob)+(1-y)*np.log(1-prob)
      loglik=-1*sum(vlog)
      return(loglik)
    
    probitmodel = minimize(probitlik, x0 = np.asarray([0.5,0.55,0.6,1]),args = (killer_app,size,price,score),method='BFGS')
    
    def hessian(x0,*args):
      epsilon=1.e-3
      f1 = approx_fprime(x0,probitlik,*args)
      n = x0.shape[0]
      hessian = np.zeros ([n,n])
      xx = x0
      for j in range(0,n):
        xx0 = xx[j] 
        xx[j] = xx0 + epsilon 
        f2 = approx_fprime(x0,probitlik,*args)
        hessian[:,j] = (f2 - f1)/epsilon
        xx[j] = xx0 
      return hessian
    
    eps = 1e-07
    hess = hessian(probitmodel.x,[eps, np.sqrt(200) * eps],[killer_app,size, price, score])
    inv_hess = np.linalg.inv(hess)
    std_beta = np.sqrt(np.diag(inv_hess))
    print(std_beta)

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-23-b78abca11ff5> in <module>()
     26 
     27 eps = 1e-07
---> 28 hess = hessian(probitmodel.x,[eps, np.sqrt(200) * eps],[killer_app,size, price, score])
     29 inv_hess = np.linalg.inv(hess)
     30 std_beta = np.sqrt(np.diag(inv_hess))

2 frames
/usr/local/lib/python3.6/dist-packages/scipy/optimize/optimize.py in _approx_fprime_helper(xk, f, epsilon, args, f0)
    689     """
    690     if f0 is None:
--> 691         f0 = f(*((xk,) + args))
    692     grad = numpy.zeros((len(xk),), float)
    693     ei = numpy.zeros((len(xk),), float)

TypeError: probitlik() missing 3 required positional arguments: 'x1', 'x2', and 'x3'
4

1 回答 1

0

很难，因为错误在第 691 行。此外，根本没有评论，而且参数的传递似乎非常复杂。开始调试，当它在第 691 行失败时，检查*((xk,) + args). 显然，只有两个值被传递，probitlik()而五个是必需的——这在您在问题中发布的最后一行中进行了解释。

于 2020-12-17T11:45:43.447 回答