作者celestialgod (攸蓝)
看板R_Language
标题Re: [问题] 回归参数的撷取
时间Wed Aug 7 14:07:08 2013
# 我造的范例
n = 4; p = 13
logit_trans = function(x) exp(x) / (1+exp(x))
data = matrix(rnorm(n*p, 0, 0.1), n)
data = data.frame(y = sapply(logit_trans(cbind(1, data) %*% 1:(p+1)),
function(p) rbinom(1,1,p)), data)
comb = t(combn(4:12,2))
beta = mapply(function(i, j) matrix(coef(glm( data[,1] ~ data[,2] + data[,3] +
data[,i] + data[,j] + data[,i]:data[,j], family=binomial(link="logit"),
data = data))), comb[,1], comb[,2])
beta = cbind(comb, t(beta))
colnames(beta) <- c("i", "j","Intercept", "Gender", "Age", "CAR1", "CAR2",
"CAR1*CAR2")
--
※ 发信站: 批踢踢实业坊(ptt.cc)
◆ From: 140.116.152.213
※ 编辑: celestialgod 来自: 140.116.152.213 (08/07 14:10)
1F：推 diplazium:虽然不知为何会有NA(singularities是啥?),不过如果只是 08/07 17:44
2F：→ diplazium:要抽出参数值的话,就直接coef(mylogit)即可~ 08/07 17:45
3F：→ celestialgod:singularities就是n<p的情况 造成无法估计 08/07 18:48
4F：→ celestialgod:原po想要完整的TABLE 08/07 18:48
5F：→ celestialgod:我查看summary.glm的code 实在无法解决QQ 08/07 18:49
6F：推 andrew43:mylogit$coefficients 08/07 23:32
※ 编辑: celestialgod 来自: 61.227.245.110 (08/08 13:08)
※ 编辑: celestialgod 来自: 61.227.245.110 (08/08 13:09)