Question

Suppose I have a data frame, df, that looks like:

f t1 t2 t3
h 1  3  4
h 2  4  3
t 3  4  5
t 5  6  8

with f being a factor and $t attributes being numerical values related to time ordered events.
I could overlay time series t1 to t3 using par(new=T) and isolate by factor manually.
But I wonder if there is some way to do this with lattice, where the overlaid time series
are conditioned by the factor. So we would have two panels, with overlaid time series corresponding to conditional factors, f. Most examples I’ve seen only use one time series (vector) per factor. I also thought about using a parallel plot, but time information is lost.
I’ve also tried something like

xyplot(df$t1+df$t2+df$t3 ~seq(3) | factor(df$f))

, but it loses row sequence connections. Anyone know if this is possible?

Here’s a very crude illustration using non lattice approach.

x<-matrix(seq(12),4,3)
f<-c('a','a','b','b')
df<-data.frame(f,x)

layout(1:2); yr<-c(0,12); xr<-c(1,3);
plot(as.numeric(df[1,2:4])~seq(3),type='o',ylim=yr,xlim=xr,ylab='A')
par(new=T)
plot(as.numeric(df[2,2:4])~seq(3),type='o',ylim=yr,xlim=xr,ylab='A')

plot(as.numeric(df[3,2:4])~seq(3), type='o',ylim=yr,xlim=xr,ylab='B')
par(new=T)
plot(as.numeric(df[4,2:4])~seq(3),type='o',ylim=yr,xlim=xr,ylab='B')

Answer 1

I added an ID variable and melted with package:reshape2

 dat
  f t1 t2 t3 ID
1 h  1  3  4  1
2 h  2  4  3  2
3 t  3  4  5  3
4 t  5  6  8  4
datm <- melt(dat, id.vars=c("ID","f"), measure.vars=c("t1", "t2", "t3"))
> datm
   ID f variable value
1   1 h       t1     1
2   2 h       t1     2
3   3 t       t1     3
4   4 t       t1     5
5   1 h       t2     3
6   2 h       t2     4
7   3 t       t2     4
8   4 t       t2     6
9   1 h       t3     4
10  2 h       t3     3
11  3 t       t3     5
12  4 t       t3     8

Since you asked to have it “overlayed” I used the group parameter to keep the ID’s separate and the “|” operator to give you the two panels for “h” and “t”:

xyplot(value~variable|f, group=ID, data=datm, type="b")