Dear R community,
I am a newbie to R and I am using lme() to analyzed a two way repeated
measures ANCOVA on some data I have gathered. In producing some
graphs based on the fixed effects I noticed that I get vary different
results depending on how I order my levels in my factor statement (see
code below). Now, I have read that different models treat data from
the factor class different for ordered vs non-ordered factors but I am
unsure as to how this works for lme(). Shown below is a set of code
used to get the coefficients for two different orders of factor levels.
THE CODE USED TO GET THE COEFFICIENTS:
#Abbr = Subject
#PDSI = covariate
#PDSI2 = PDSI^2 (To fit a polynomial)
#Set the Level order
i.levels<-c("Mixed","Short","Tall","NoType")
#I am not interested in this level for my comparison
data.sub <- data[which(data$indep != "NoType"),]
#Convert the data into factors
data$indep<-factor(data$indep,
levels = i.levels)
#Run the model
F <- lme(response~indep+PDSI+PDSI2+indep:PDSI+indep:PDSI2,
random = ~PDSI|Abbr, data = data.sub, method = "ML")
#Get the coefficients for the fixed effects
summary(F)$coefficients$fixed
THE OUTCOME:
> i.levels<-c("Tall","Short","Mixed","NoType")
> summary(F)$coefficients$fixed
(Intercept) indep1 indep2 PDSI PDSI2
indep1:PDSI
115.20850196 2.13074177 -18.64650516 2.32042307 -0.46510460
0.09965080
indep2:PDSI indep1:PDSI2 indep2:PDSI2
0.90607575 0.01106839 -0.28827352
Other order:
> i.levels<-c("Mixed","Short","Tall","NoType")
> summary(F)$coefficients$fixed
(Intercept) indep1 indep2 PDSI PDSI2
indep1:PDSI
115.2085020 16.5157634 -18.6465052 2.3204231 -0.4651046
-1.0057266
indep2:PDSI indep1:PDSI2 indep2:PDSI2
0.9060757 0.2772051 -0.2882735
Also, as I am writing this I found another peculiarity. The order of
the first declaration of the levels also matters somehow. To get the
above results, I ran a program that ran the first order of i.levels
and immediately following this code is a duplicate code with the
second order of i.levels specified. I ran them both at the same
time. When I ran the code with Mixed, Short, Tall, NoType first, both
orders produced the same coefficients. But if I ran the
Tall,Short,Mixed,NoType first then I got the above results. I am even
more lost now. Any insight would be greatly appreciated.
Thanks
Jarrod
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
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.
