在逻辑回归的glm函数中，似然函数存储在哪里？ 有家人吗

Question

I am currently trying to run a logistic regression on my own, using the functions optim , nlm , etc. However, I am somehow getting different coefficient values for the custom model as compared to when I run glm.

I suspect that my likelihood function might be different, but cannot find where the logit likelihood function is stored in glm . Does anyone know?

Answer 1

The likelihood function is not stored in the glm object. In R 3.1, only the following fields are present:

> m <- glm(y ~ x, data=df, family=binomial(link=logit))
> names(m)
 [1] "coefficients"      "residuals"         "fitted.values"    
 [4] "effects"           "R"                 "rank"             
 [7] "qr"                "family"            "linear.predictors"
[10] "deviance"          "aic"               "null.deviance"    
[13] "iter"              "weights"           "prior.weights"    
[16] "df.residual"       "df.null"           "y"                
[19] "converged"         "boundary"          "model"            
[22] "call"              "formula"           "terms"            
[25] "data"              "offset"            "control"          
[28] "method"            "contrasts"         "xlevels"

If you type glm.fit at the R prompt, you will see how the optimization proceeds. In particular, there is no separate likelihood function that can be evaluated with different points in the parameter space. See the link in Steve's comment for what you need to do.