1

folks. I try to get the standard errors of coefficients manually from probit regression analysis in R. These are 0.69012 and 0.03565 from the following R program.

> binary_data
   id X1 Y
1   1 14 0
2   2 29 0
3   3  6 0
4   4 25 1
5   5 18 1
6   6  4 0
7   7 18 0
8   8 12 0
9   9 22 1
10 10  6 0
11 11 30 1
12 12 11 0
13 13 30 1
14 14  5 0
15 15 20 1
16 16 13 0
17 17  9 0
18 18 32 1
19 19 24 0
20 20 13 1
21 21 19 0
22 22  4 0
23 23 28 1
24 24 22 1
25 25  8 1
> probit_output <- glm(Y ~ X1, family=binomial(link="probit"), data=binary_data)
> summary(probit_output)

Call:
glm(formula = Y ~ X1, family = binomial(link = "probit"), data = binary_data)


Deviance Residuals: 
    Min       1Q   Median       3Q      Max

-1.8959  -0.7579  -0.3907   0.8101   1.9691


Coefficients:
            Estimate Std. Error z value Pr(>|z|)

(Intercept) -1.83787    0.69012  -2.663  0.00774 **
X1           0.09686    0.03565   2.717  0.00659 **
Shayan Shafiq
  • 643
Sangwoo.Statistics
  • 11
  • Is confint(probit_output) what you are looking for? Doing it manually is complicated, because it is based n the asymptotic $\chi^2$ distribution of likelihood-ratios. You can check the implementation in R with getFromNameSpace("confint.hlm", "MASS") and getFromNameSpace("profile.hlm", "MASS"). – cdalitz Dec 14 '21 at 06:57

0 Answers0