Question

我需要使用R修复逻辑回归模型。这是Stata代码：

melogit num, binomial(varsum)

我在R中尝试代码但结果却不同。这是我在R的代码。

summary(glm(cbind(num,nonum) ~ -1 + varsum, family = binomial("logit")))

Stata的输出

Logistic regression                             Number of obs     =         18
Binomial variable:         vsum

                                                Wald chi2(0)      =          .
Log likelihood = -26.242541                     Prob > chi2       =          .
------------------------------------------------------------------------------
         num |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |  -1.170071   .2202982    -5.31   0.000    -1.601848   -.7382947
------------------------------------------------------------------------------

来自R

Call:
glm(formula = cbind(num, nonum) ~ vsum - 1, family = binomial("logit"))

Deviance Residuals: 
     Min        1Q    Median        3Q       Max  
-13.5137   -3.9972   -0.7592    2.8821   10.7677  

Coefficients:
   Estimate Std. Error z value Pr(>|z|)    
vsum -0.82854    0.03839  -21.58   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 3198.67  on 18  degrees of freedom
Residual deviance:  673.14  on 17  degrees of freedom
AIC: 704.92

Number of Fisher Scoring iterations: 7

这些是数据：

num = c(0,1,2,0,5,1,1,1,1,1,0,0,3,6,0,0,1,4) 

nonum = c(116,43,206,130,146,97,173,73,96,112,66,70,185,181,118,252,344,60) 

varsum = c(3,4,11,7,11,5,4,3,20,3,7,1,8,17,0,1,4,5)

Answer 1

你适合两种不同的模型。

首先，melogit定义了一个多级混合效应逻辑回归模型，而glm(..., family = binomial("logit"))符合一个简单的逻辑回归模型。

我不熟悉Stata，但快速搜索表明，可以使用glm结合link(logit) family(binomial) robust选项估算具有比例数据的等效逻辑回归模型。

将Stata代码更改为R.

1 个答案: