On 12/30/18 5:31 PM, Marc Girondot via R-help wrote:
Dear members,
Let do a example of simple GLMM with x and G as fixed factors and R as
random factor:
(note that question is the same with GLM or even LM):
x <- rnorm(100)
y <- rnorm(100)
G <- as.factor(sample(c("A", "B", "C", "D"), 100, replace = TRUE))
R <- as.factor(rep(1:25, 4))
library(lme4)
m <- lmer(y ~ x + G + (1 | R))
summary(m)$coefficients
I get the fixed effect fit and their SE
> summary(m)$coefficients
Estimate Std. Error t value
(Intercept) 0.07264454 0.1952380 0.3720820
x -0.02519892 0.1238621 -0.2034433
GB 0.10969225 0.3118371 0.3517614
GC -0.09771555 0.2705523 -0.3611706
GD -0.12944760 0.2740012 -0.4724344
The estimate for GA is not shown as it is fixed to 0. Normal, it is the
reference level.
But is there a way to get SE for GA of is-it non-sense question because
GA is fixed to 0 ?
In a way, yes it's a nonsense question, as you say.
If you really want an SE for GA then re-parametrise so that GA is
meaningful:
m2 <- lmer(y ~ x + 0 + G + (1 | R))
Note that with this formulation GA will be there, "(Intercept)" will
disappear, and GB, GC and GD will now mean something different.
GA from m2 = (Intercept) from m
GB from m2 = (Intercept) + GB from m
GC from m2 = (Intercept) + GC from m
GD from m2 = (Intercept) + GD from m
I haven't followed what you've done below, but I think that you are
making things unnecessarily complicated and life difficult for yourself.
cheers,
Rolf
--
Technical Editor ANZJS
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276
______________
I propose here a solution but I don't know if it is correct. It is based
on reordering levels and averaging se for all reordering:
G <- relevel(G, "A")
m <- lmer(y ~ x + G + (1 | R))
sA <- summary(m)$coefficients
G <- relevel(G, "B")
m <- lmer(y ~ x + G + (1 | R))
sB <- summary(m)$coefficients
G <- relevel(G, "C")
m <- lmer(y ~ x + G + (1 | R))
sC <- summary(m)$coefficients
G <- relevel(G, "D")
m <- lmer(y ~ x + G + (1 | R))
sD <- summary(m)$coefficients
seA <- mean(sB["GA", "Std. Error"], sC["GA", "Std. Error"], sD["GA",
"Std. Error"])
seB <- mean(sA["GB", "Std. Error"], sC["GB", "Std. Error"], sD["GB",
"Std. Error"])
seC <- mean(sA["GC", "Std. Error"], sB["GC", "Std. Error"], sD["GC",
"Std. Error"])
seD <- mean(sA["GD", "Std. Error"], sB["GD", "Std. Error"], sC["GD",
"Std. Error"])
seA; seB; seC; seD
Thanks,
Marc
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
