R help - Oct 2009 - problem understanding factor levels for use lattice panel order

Hello all,
In order to get the desired order of panels within lattice I realized I needed
to define my factor levels.
I now see how to properly do it (attached code), but:
1. I don't understand why I have to use factor() on a variable that is
already a factor, and why levels() alone doesn't work to readjust the level
order.
2. I just noticed that if I do use levels(xx$v1), it's changing something
that can't be repaired with:
 xx$v1<-factor(xx$v1,levels=c('b','a','c')), and both
plots are wrong.
I'm failing to grasp what levels() is doing compared to factor().
thanks very much.
Michael Folkes
 
#__________________________________
 
xx<-data.frame(v1=rep(letters[1:3],rep(3,3)),v2=rep(1:3,3),v3=rep(1:3,rep(3,3)))
str(xx)
levels(xx$v1)
 
#INCORRECT:
#if I run this level(), the factor() below doens't work properly
#levels(xx$v1)<-c('b','a','c')
xx
g1<-xyplot(v3~v2|v1,data=xx)

#CORRECT:
xx$v1<-factor(xx$v1,levels=c('b','a','c'))
g2<-xyplot(v3~v2|v1,data=xx)
xx
print(g1,split=c(1,1,2,1),more=T)
print(g2,split=c(2,1,2,1),more=F)

 
 
_______________________________________________________
Michael Folkes
Salmon Stock Assessment
Canadian Dept. of Fisheries & Oceans     
Pacific Biological Station
3190 Hammond Bay Rd.
Nanaimo, B.C., Canada
V9T-6N7
Ph (250) 756-7264 Fax (250) 756-7053  Michael.Folkes@dfo-mpo.gc.ca
<mailto:Michael.Folkes@dfo-mpo.gc.ca>

	[[alternative HTML version deleted]]

On 9/10/2009, at 6:52 AM, Folkes, Michael wrote:
> Hello all,
> In order to get the desired order of panels within lattice I  
> realized I needed to define my factor levels.
> I now see how to properly do it (attached code), but:
> 1. I don't understand why I have to use factor() on a variable that  
> is already a factor, and why levels() alone doesn't work to  
> readjust the level order.
> 2. I just noticed that if I do use levels(xx$v1), it's changing  
> something that can't be repaired with:
>  xx$v1<-factor(xx$v1,levels=c('b','a','c')), and
both plots are wrong.
> I'm failing to grasp what levels() is doing compared to factor().
> thanks very much.
> Michael Folkes
>
> #__________________________________
>
> xx<-data.frame(v1=rep(letters[1:3],rep(3,3)),v2=rep(1:3,3),v3=rep 
> (1:3,rep(3,3)))
> str(xx)
> levels(xx$v1)
>
> #INCORRECT:
> #if I run this level(), the factor() below doens't work properly
> #levels(xx$v1)<-c('b','a','c')
> xx
> g1<-xyplot(v3~v2|v1,data=xx)
>
> #CORRECT:
> xx$v1<-factor(xx$v1,levels=c('b','a','c'))
> g2<-xyplot(v3~v2|v1,data=xx)
> xx
> print(g1,split=c(1,1,2,1),more=T)
> print(g2,split=c(2,1,2,1),more=F)
The object xx$v1 is ``really'' the vector of indices 1 1 1 2 2 2 3 3 3
Initially it has a levels attribute c('a','b','c') which
says to
associate the symbol 'a' with index 1, 'b', with index 2, and
'c'
with index 3.
When you type xx$v1 you see the sequence a a a b b b c c c.

The call

	levels(xx$v1) <- c('b','a','c')

says to associate the symbol 'b' with index 1, 'a', with index
2, and
'c' with index 3.
Now when you type xx$v1 you see the sequence b b b a a a c c c.

The call

	xx$v1 <- factor(xx$v1,levels=c('b','a','c'))

(done ***instead*** of the the call levels(xx$v1) <-
c('b','a','c'))

essentially replaces xx$v1 by
match(xx$v1,c('b','a','c')) so that now
xx$v1 is
``really'' the vector of indices 2 2 2 1 1 1 3 3 3.  The levels  
attribute says
that 'b' is associated with 1, 'a' is associated with 2, and
'c' is
associated with 3.
So when you type xx$v1 you see a a a b b b c c c as before.

	cheers,

		Rolf Turner

######################################################################
Attention:\ This e-mail message is privileged and confid...{{dropped:9}}