I have a variable yi
that represents a treatment effect over time nyears
for a bunch of different studies ( Site
). There are also two grouping factors with two levels each: N
(Nhigh/Nlow) and Myc
(AM/ECM). I need to know if yi
shows a significant positive or negative trend over time nyears
, and if the trends changes among subgroups N
x Myc
.
The mixed-effects models shows a significant triple interaction nyears
* N
* Myc
library(lme4)
library(car)
> mod <- lmer(yi ~ N*Myc*nyears + (1|Site), data = df)
> Anova(mod)
Analysis of Deviance Table (Type II Wald chisquare tests)
Response: yi
Chisq Df Pr(>Chisq)
N 0.7468 1 0.387489
Myc 0.0875 1 0.767403
nyears 1.1217 1 0.289559
N:Myc 0.5428 1 0.461272
N:nyears 2.2371 1 0.134733
Myc:nyears 0.6318 1 0.426691
N:Myc:nyears 10.8108 1 0.001009 **
How can I now find out the sign of the slope and significance for each of the 4 subgroups?
Thanks
The value of the slope (on nyears
, right?) is given by
nyears
nyears + N:nyears
nyears + Myc:nyears
nyears + Myc:nyears + N:Myc:nyears
for the four respective groups. (Are N
and Myc
numeric 0/1? They don't look like factors, judging by the output. If they aren't 0/1 the recode.)
For significance testing of the slopes, either use linearHypothesis
in the CAR
package; or you could use eg waldtest
in package lmtest
; or rewrite your model so that nyears
is the coefficient of interest for each of the four groups. (For example, set Myc
equal to 1 minus its old value, if it was a 0/1 dummie.)
I would use the nlme
package to code the mixed effects model and then check the output using summary
. This will report slopes, signs and p values.
require(nlme)
m1<- lme(yi ~ N*Myc*nyears, random= ~1|Site, data=df)
summary(m1)
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.