R help - Feb 2010 - xyplot 3 panels 3 different Y variables

Often, when exploring a dataset, I'd like to plot several very different Y
variables against the same X variable, in panels stacked one over the other. Is
there an easy way to do this?

I'd like to achieve an elegant look similar to the look achieved by lattice
in conditioned plots--for instance no space between panels. But unlike in
straightforward conditioned plot, each panel may be on a different scale.

Example.

 	?	Plot Estrogen, Creatinine, and their ratio; all by the same predictor
variable (say, Day).

 		Or: In a longitudinal study of hormones in reproductive-age women, plot
progesterone, estradiol, testosterone, luteinizing hormone, follicle-stimulating
hormone, and thyroid-stimulating hormone all on one page, parallel. Note that
several of these variables are measured in different units.

 	?	One panel for each outcome variable, arranged one above the other.

 	?	Minimal vertical space between the panels.

To make this concrete, let's generate toy data:

N<-40
set.seed(5234767)
JUNK<-data.frame(Day=1:N)
JUNK$Creatinine<-exp(2*rnorm(nrow(JUNK)))
JUNK$Estrogen<- (sin(JUNK$Day/pi) + 2.5) * ( exp(2*rnorm(nrow(JUNK))) *
JUNK$Creatinine )
JUNK$Creatinine[10]<-0.0001
JUNK$Ratio<- JUNK$Estrogen / JUNK$Creatinine

The following traditional graphics commands put an annoying wide space between
the "panels" by default. Also, the X ticks are repeated unnecessarily.

par(mfrow=c(3,1))
par(oma=c(0,0,1,0))
plot(JUNK$Day, JUNK$Estrogen, xlab="", ylab="Estrogen",
type="o")
plot(JUNK$Day, JUNK$Creatinine, xlab="", ylab="Creatinine",
type="o")
plot(JUNK$Day, JUNK$Ratio, xlab="Day", ylab="Ratio",
type="o")

The following lattice approach gives a kinda nice-looking end product, but seems
so counterintuitive that I want to call it a workaround.  For instance, it
generates a "time" variable that actually is a category.  And the
variable names are converted into the levels of a factor by hand, so that the
process is susceptible to human error.

Also, the Y variables are not labeled on the axes, only in the strip.  This is
not ideal.

JUNK2<-JUNK
names(JUNK2)[names(JUNK2)=="Creatinine"]<-"Y.1"
names(JUNK2)[names(JUNK2)=="Estrogen"]<-"Y.2"
names(JUNK2)[names(JUNK2)=="Ratio"]<-"Y.3"
JUNKlong<- reshape(JUNK2,  dir="long", varying=2:4)
JUNKlong$outcome<-factor( JUNKlong$time, levels=1:3,
labels=c("Creatinine", "Estrogen", "Ratio") )
JUNKlong$time<-NULL
library(lattice)
xyplot( Y ~ Day | outcome, data=JUNKlong, layout=c(1,3), type="o",
scales=list(x=list(alternating=3), y=list(relation="free",
alternating=3)), ylab="")

Am I making this harder than it needs to be?

Thanks

Jacob A. Wegelin
Assistant Professor
Department of Biostatistics
Virginia Commonwealth University
730 East Broad Street Room 3006
P. O. Box 980032
Richmond VA 23298-0032
U.S.A.

As these are time series you can use plot.ts, plot.zoo or xyplot.zoo:

tt <- ts(JUNK[-1])
plot(tt)

library(zoo)
z <- zoo(JUNK[-1], JUNK[,1])
plot(z)

library(lattice)
xyplot(z)


On Thu, Feb 4, 2010 at 5:30 PM, Jacob Wegelin <jacobwegelin at
fastmail.fm> wrote:
>
> Often, when exploring a dataset, I'd like to plot several very
different Y
> variables against the same X variable, in panels stacked one over the
other.
> Is there an easy way to do this?
>
> I'd like to achieve an elegant look similar to the look achieved by
lattice
> in conditioned plots--for instance no space between panels. But unlike in
> straightforward conditioned plot, each panel may be on a different scale.
>
> Example.
>
> ? ? ? ?? ? ? ? Plot Estrogen, Creatinine, and their ratio; all by the same
> predictor variable (say, Day).
>
> ? ? ? ? ? ? ? ?Or: In a longitudinal study of hormones in reproductive-age
> women, plot progesterone, estradiol, testosterone, luteinizing hormone,
> follicle-stimulating hormone, and thyroid-stimulating hormone all on one
> page, parallel. Note that several of these variables are measured in
> different units.
>
> ? ? ? ?? ? ? ? One panel for each outcome variable, arranged one above the
> other.
>
> ? ? ? ?? ? ? ? Minimal vertical space between the panels.
>
> To make this concrete, let's generate toy data:
>
> N<-40
> set.seed(5234767)
> JUNK<-data.frame(Day=1:N)
> JUNK$Creatinine<-exp(2*rnorm(nrow(JUNK)))
> JUNK$Estrogen<- (sin(JUNK$Day/pi) + 2.5) * ( exp(2*rnorm(nrow(JUNK))) *
> JUNK$Creatinine )
> JUNK$Creatinine[10]<-0.0001
> JUNK$Ratio<- JUNK$Estrogen / JUNK$Creatinine
>
> The following traditional graphics commands put an annoying wide space
> between the "panels" by default. Also, the X ticks are repeated
> unnecessarily.
>
> par(mfrow=c(3,1))
> par(oma=c(0,0,1,0))
> plot(JUNK$Day, JUNK$Estrogen, xlab="", ylab="Estrogen",
type="o")
> plot(JUNK$Day, JUNK$Creatinine, xlab="",
ylab="Creatinine", type="o")
> plot(JUNK$Day, JUNK$Ratio, xlab="Day", ylab="Ratio",
type="o")
>
> The following lattice approach gives a kinda nice-looking end product, but
> seems so counterintuitive that I want to call it a workaround. ?For
> instance, it generates a "time" variable that actually is a
category. ?And
> the variable names are converted into the levels of a factor by hand, so
> that the process is susceptible to human error.
>
> Also, the Y variables are not labeled on the axes, only in the strip. ?This
> is not ideal.
>
> JUNK2<-JUNK
> names(JUNK2)[names(JUNK2)=="Creatinine"]<-"Y.1"
> names(JUNK2)[names(JUNK2)=="Estrogen"]<-"Y.2"
> names(JUNK2)[names(JUNK2)=="Ratio"]<-"Y.3"
> JUNKlong<- reshape(JUNK2, ?dir="long", varying=2:4)
> JUNKlong$outcome<-factor( JUNKlong$time, levels=1:3,
labels=c("Creatinine",
> "Estrogen", "Ratio") )
> JUNKlong$time<-NULL
> library(lattice)
> xyplot( Y ~ Day | outcome, data=JUNKlong, layout=c(1,3),
type="o",
> scales=list(x=list(alternating=3), y=list(relation="free",
alternating=3)),
> ylab="")
>
> Am I making this harder than it needs to be?
>
> Thanks
>
> Jacob A. Wegelin
> Assistant Professor
> Department of Biostatistics
> Virginia Commonwealth University
> 730 East Broad Street Room 3006
> P. O. Box 980032
> Richmond VA 23298-0032
> U.S.A.
> ______________________________________________
> R-help at 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.
>
>

Hi:

Here's a ggplot2 solution for your consideration:

Start by melting the data using package reshape; this stacks the response
variables and uses their names as factor levels in a conditioning plot:

library(ggplot2)     # loads plyr and reshape in the process

junk <- melt(JUNK, id = "Day")
dim(junk)
# [1] 120   3
head(junk, n = 3)
  Day   variable       value
1   1 Creatinine  2.81193872
2   2 Creatinine  3.71031330
3   3 Creatinine  9.36436091

p <- ggplot(junk, aes(x = Day, y = value))
p + geom_line() + facet_wrap(~ variable, scale = 'free_y', nrow =3)

HTH,
Dennis

On Thu, Feb 4, 2010 at 2:30 PM, Jacob Wegelin
<jacobwegelin@fastmail.fm>wrote:
>
> Often, when exploring a dataset, I'd like to plot several very
different Y
> variables against the same X variable, in panels stacked one over the
other.
> Is there an easy way to do this?
>
> I'd like to achieve an elegant look similar to the look achieved by
lattice
> in conditioned plots--for instance no space between panels. But unlike in
> straightforward conditioned plot, each panel may be on a different scale.
>
> Example.
>
>        •       Plot Estrogen, Creatinine, and their ratio; all by the same
> predictor variable (say, Day).
>
>                Or: In a longitudinal study of hormones in reproductive-age
> women, plot progesterone, estradiol, testosterone, luteinizing hormone,
> follicle-stimulating hormone, and thyroid-stimulating hormone all on one
> page, parallel. Note that several of these variables are measured in
> different units.
>
>        •       One panel for each outcome variable, arranged one above the
> other.
>
>        •       Minimal vertical space between the panels.
>
> To make this concrete, let's generate toy data:
>
> N<-40
> set.seed(5234767)
> JUNK<-data.frame(Day=1:N)
> JUNK$Creatinine<-exp(2*rnorm(nrow(JUNK)))
> JUNK$Estrogen<- (sin(JUNK$Day/pi) + 2.5) * ( exp(2*rnorm(nrow(JUNK))) *
> JUNK$Creatinine )
> JUNK$Creatinine[10]<-0.0001
> JUNK$Ratio<- JUNK$Estrogen / JUNK$Creatinine
>
> The following traditional graphics commands put an annoying wide space
> between the "panels" by default. Also, the X ticks are repeated
> unnecessarily.
>
> par(mfrow=c(3,1))
> par(oma=c(0,0,1,0))
> plot(JUNK$Day, JUNK$Estrogen, xlab="", ylab="Estrogen",
type="o")
> plot(JUNK$Day, JUNK$Creatinine, xlab="",
ylab="Creatinine", type="o")
> plot(JUNK$Day, JUNK$Ratio, xlab="Day", ylab="Ratio",
type="o")
>
> The following lattice approach gives a kinda nice-looking end product, but
> seems so counterintuitive that I want to call it a workaround.  For
> instance, it generates a "time" variable that actually is a
category.  And
> the variable names are converted into the levels of a factor by hand, so
> that the process is susceptible to human error.
>
> Also, the Y variables are not labeled on the axes, only in the strip.  This
> is not ideal.
>
> JUNK2<-JUNK
> names(JUNK2)[names(JUNK2)=="Creatinine"]<-"Y.1"
> names(JUNK2)[names(JUNK2)=="Estrogen"]<-"Y.2"
> names(JUNK2)[names(JUNK2)=="Ratio"]<-"Y.3"
> JUNKlong<- reshape(JUNK2,  dir="long", varying=2:4)
> JUNKlong$outcome<-factor( JUNKlong$time, levels=1:3,
labels=c("Creatinine",
> "Estrogen", "Ratio") )
> JUNKlong$time<-NULL
> library(lattice)
> xyplot( Y ~ Day | outcome, data=JUNKlong, layout=c(1,3),
type="o",
> scales=list(x=list(alternating=3), y=list(relation="free",
alternating=3)),
> ylab="")
>
> Am I making this harder than it needs to be?
>
> Thanks
>
> Jacob A. Wegelin
> Assistant Professor
> Department of Biostatistics
> Virginia Commonwealth University
> 730 East Broad Street Room 3006
> P. O. Box 980032
> Richmond VA 23298-0032
> U.S.A.
> ______________________________________________
> 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.
>
>
	[[alternative HTML version deleted]]

On Thu, Feb 4, 2010 at 2:30 PM, Jacob Wegelin <jacobwegelin at
fastmail.fm> wrote:
>
> Often, when exploring a dataset, I'd like to plot several very
different Y
> variables against the same X variable, in panels stacked one over the
other.
> Is there an easy way to do this?
>
> I'd like to achieve an elegant look similar to the look achieved by
lattice
> in conditioned plots--for instance no space between panels. But unlike in
> straightforward conditioned plot, each panel may be on a different scale.
>
> Example.
>
> ? ? ? ?? ? ? ? Plot Estrogen, Creatinine, and their ratio; all by the same
> predictor variable (say, Day).
>
> ? ? ? ? ? ? ? ?Or: In a longitudinal study of hormones in reproductive-age
> women, plot progesterone, estradiol, testosterone, luteinizing hormone,
> follicle-stimulating hormone, and thyroid-stimulating hormone all on one
> page, parallel. Note that several of these variables are measured in
> different units.
>
> ? ? ? ?? ? ? ? One panel for each outcome variable, arranged one above the
> other.
>
> ? ? ? ?? ? ? ? Minimal vertical space between the panels.
>
> To make this concrete, let's generate toy data:
>
> N<-40
> set.seed(5234767)
> JUNK<-data.frame(Day=1:N)
> JUNK$Creatinine<-exp(2*rnorm(nrow(JUNK)))
> JUNK$Estrogen<- (sin(JUNK$Day/pi) + 2.5) * ( exp(2*rnorm(nrow(JUNK))) *
> JUNK$Creatinine )
> JUNK$Creatinine[10]<-0.0001
> JUNK$Ratio<- JUNK$Estrogen / JUNK$Creatinine
>
> The following traditional graphics commands put an annoying wide space
> between the "panels" by default. Also, the X ticks are repeated
> unnecessarily.
>
> par(mfrow=c(3,1))
> par(oma=c(0,0,1,0))
> plot(JUNK$Day, JUNK$Estrogen, xlab="", ylab="Estrogen",
type="o")
> plot(JUNK$Day, JUNK$Creatinine, xlab="",
ylab="Creatinine", type="o")
> plot(JUNK$Day, JUNK$Ratio, xlab="Day", ylab="Ratio",
type="o")
>
> The following lattice approach gives a kinda nice-looking end product, but
> seems so counterintuitive that I want to call it a workaround. ?For
> instance, it generates a "time" variable that actually is a
category. ?And
> the variable names are converted into the levels of a factor by hand, so
> that the process is susceptible to human error.
>
> Also, the Y variables are not labeled on the axes, only in the strip. ?This
> is not ideal.
>
> JUNK2<-JUNK
> names(JUNK2)[names(JUNK2)=="Creatinine"]<-"Y.1"
> names(JUNK2)[names(JUNK2)=="Estrogen"]<-"Y.2"
> names(JUNK2)[names(JUNK2)=="Ratio"]<-"Y.3"
> JUNKlong<- reshape(JUNK2, ?dir="long", varying=2:4)
> JUNKlong$outcome<-factor( JUNKlong$time, levels=1:3,
labels=c("Creatinine",
> "Estrogen", "Ratio") )
> JUNKlong$time<-NULL
> library(lattice)
> xyplot( Y ~ Day | outcome, data=JUNKlong, layout=c(1,3),
type="o",
> scales=list(x=list(alternating=3), y=list(relation="free",
alternating=3)),
> ylab="")
>
> Am I making this harder than it needs to be?
Not really, but there is some simplification to be gained using the
`extended' formula notation which does the reshaping internally.

xyplot(Creatinine + Estrogen + Ratio ~ Day, data=JUNK, outer = TRUE,
       layout = c(1,3), type="o", ylab = NULL,
       scales = list(x=list(alternating=3),
                     y=list(relation="free", rot = 0)))


xyplot(Creatinine + Estrogen + Ratio ~ Day, data=JUNK, outer = TRUE,
       layout = c(1,3), type="o", strip = FALSE,
       ylab = c("Creatinine", "Estrogen",
"Ratio"),
       scales = list(x=list(alternating=3),
                     y=list(relation="free", rot = 0)))


-Deepayan

Continuing from your JUNK example:

xyplot(Creatinine + Estrogen + Ratio ~ Day, data=JUNK, type="l",
outer=TRUE, layout=c(1,3),
         scales=list(y=list(relation="free")))