1

我正在y ~ v + m + s + m:s + (1|subunit)用 lmer() 拟合模型。 s是一个与 相互作用的连续变量m，一个具有 3 个级别的分类因子：A、B和C。

拟合模型使用A作为因子的参考水平m：

fit_ref_A <- lmer(y ~ 1 + v + m*s + (1|subunit), data=df)

B和C的参数估计值可以使用 A 的估计值以及B和C的差值来计算。我有兴趣提取置信区间。

运行 confint() 给出As处变量斜率的置信区间。我还对B和C的斜率的置信区间感兴趣，而不仅仅是与A的斜率差异的置信区间。有没有办法从中提取这个？到目前为止，我唯一能想到的是以B为参考重新调平，适合新的配合，然后以C作为参考重新调平，并适合第三次配合。sfit_ref_A

问题：有没有办法从中提取所有内容（尤其是置信区间）fit_ref_A？

代码：

library(lme4)
# create the dataset, unbalanced at the lowest stratum ( 2 repeats for m==A instead of 3)
set.seed(2)
s_levels <- 1:5
m_levels <- c("A", "B", "C")
v_levels <- c("L2", "L3", "L4")
reps <- 1:3
df <- expand.grid(rep=reps, s=s_levels, m=m_levels, v=v_levels)
df$subunit <- as.factor(paste(df$v,"-",df$m,"-",df$s, sep=""))
df$y <- rnorm(nrow(df), 0, 1)
df <- subset(df, !(rep==3 & m=="A"))  # drop the 3rd repeat for m=="A"
table(df$m)  # shows 30 for A, 45 for B, 45 for C as expected

# fit 3 different models, with three different reference levels for 'm'
fit_ref_A <- lmer(y ~ 1 + v + m*s + (1|subunit), data=df)

df$m <- relevel(df$m, ref = "B")
fit_ref_B <- lmer(y ~ 1 + v + m*s + (1|subunit), data=df)

df$m <- relevel(df$m, ref = "C")
fit_ref_C <- lmer(y ~ 1 + v + m*s + (1|subunit), data=df)

# Calculate the confidence intervals for the continuous variable s at the three
# different levels for categorical factor m.  Must use 3 separate fits.

cis_at_m_reference_A <- confint(fit_ref_A)
cis_at_m_reference_B <- confint(fit_ref_B)
cis_at_m_reference_C <- confint(fit_ref_C)

cis_at_m_reference_A["s",]
cis_at_m_reference_B["s",]
cis_at_m_reference_C["s",]

# Any way to just extract all three from fit_ref_A?
4

1 回答 1

2

您可以通过高斯误差传播获得近似 CI：

sum(fixef(fit_ref_A)[c("s", "mB:s")]) + 
  c(-1.96, 1.96) * sqrt(sum(vcov(fit_ref_A)[c("s", "mB:s"), c("s", "mB:s")]))
#[1] -0.3346310  0.1863014

或者你可以引导：

myboot <- bootMer(fit_ref_A, function(x) {
  cf <- fixef(x)
  c(sA = cf[["s"]], sB = cf[["s"]] + cf[["mB:s"]], sC = cf[["s"]] + cf[["mC:s"]])
}, nsim = 1e4, seed = 42)

apply(myboot$t, 2, quantile, probs = c(0.025, 0.975))
#              sA         sB         sC
#2.5%  -0.4022927 -0.3415690 -0.3969831
#97.5%  0.2041610  0.1858731  0.1266355
于 2020-12-23T08:03:58.100 回答