[R-sig-ME] Subset data problems with mm

CL Pressland Kate.Pressland at bristol.ac.uk
Mon Apr 27 12:12:01 CEST 2009


Dear all,

Thank for your previous comments and help on my queries regarding mixed 
models.

I now have a problem regarding subsetting my data. For previous 
explanations on how my data is structured please see post "model definition 
with repeat measures". Briefly, my data is on lepidoptera surveyed over 
weeks and years on various sites that are all under different management. I 
am interested in whether management and leps are correlated in some way. I 
would like to subset the data frame but I am having problems and I think 
this may be down to my variable definitions.

An example model I tried:

	model1<-lme(Lep~MAN, random=~1|Site, data=new.ALL, subset = (Week > 15))

I get this error message:

	Error in sprintf(gettext(fmt, domain = domain), ...) :
  		object "Call" not found

I have created the variable Week by using this code:
	Week<-ordered(Weeks_rec)

Is this the cause of my problem? I think it's important that my variables 
are labelled (e.g. MAN<-factor(management)) but does this then interfere 
with subsetting code? As I have labeled Week as ordered(Weeks_rec), would I 
have to write individually != 1 & 1= 2...& != 15? This seem quite laborious!

My data is unbalanced so I also have included the command:
	new.ALL<-na.exclude(ALL)
	names(new.ALL)
	attach(new.ALL)

I have tried creating a new 'dataset':
	new.All1<-subset(new.ALL, Week > 15)

and writing that into the model instead
	model1<-lme(Lep~MAN, random=~1|Site, data=new.ALL1)

but I still get the same error message or the model runs with all the Week 
and not the subset. Would I need to write
	subset = (Week 1:16, >15) or similar?

I am very confused as I've used subsetting before without a problem. I've 
tried as many combinations as I could think of and it still won't work. Can 
anybody shed some light on this? Full code and variables labeling below if 
this is useful.

Your help is appreciated.

Kate

-------------

#full code:

ALL<-read.csv("location\\filename.txt",header=T)

attach(ALL)
names(ALL)
library(nlme)
library(lattice)

new.ALL<-na.exclude(ALL) # allowing NAs to be ignored but still keep the 
residuals and predictions padding to the correct length of the dataset
names(new.ALL)
attach(new.ALL)

#variables with correct labeling
Site<-factor(Site_ID)
Yr<-factor(Year)
Week<-ordered(Weeks_rec)		#1, 2, 3, ... , 26
MAN<-ordered(management_code)		#0,1,2
Lep<-log(Lep_m)				#log transformed
BAP<-factor(UK_BAP)			#TRUE, FALSE
Splist<-factor(Gen_Spec)		#NA, 0 (Generalist), 1 (Specialist)
Mgrnt<-factor(Migrant)			#TRUE, FALSE
Type<-factor(Lep_type)			#butterfly, moth
Sun<-asin(sqrt(Mean_Sun/100)) 	#arcsin transformation of %
Temp<-Mean_Temp
Wind<-ordered(Mean_Wind)		#NA,0,1,2,3,4,5,6




More information about the R-sig-mixed-models mailing list