R help - Oct 2012 - rms plot.Predict question: swapping x- and y- axis for categorical predictors

Hello all,

I'm trying to plot the effects of variables estimated by a regression model
fit individually, and for categorical predictors, the independent variable
shows up on the y-axis, with the dependent variable on the x-axis. Is there
a way to prevent this reversal?

Sample code with dummy data:

# make dummy data
set.seed(1)
x1 <- runif(200)
x2 <- sample(c(1,2),200, TRUE)
x3 <- sample(c(0,1),200,T)
x4 <- runif(200)

# the dependent variable:
distance <- (x1/3 + x2 + rnorm(200)^2 - x3 - x4/2)

# factor two vars, and add to datadist:
x3 <- factor(x3)
x2 <- factor(x2)

d <- datadist(x1,x2,x3,x4)
options(datadist="d")

# Make a simple model:
f <- ols(distance ~ x1 + x2 + x4+ x3, x=T)

# plot variable effect of a categorical variable:
plot(Predict(f, x2))

^ above step generates a plot with x2 on the y-axis and distance on the
x-axis, which is the opposite of what I'm aiming for. The continuous
variables do not have this problem; nor does the plot(Predict(f)) function
to plot all of the effects at once.

Thank you so much in advance for your replies! My apologies if this question
has been answered already; I've tried searching to no avail.

Best,
Stephanie

(Stanford University, Department of Linguistics)



--
View this message in context:
http://r.789695.n4.nabble.com/rms-plot-Predict-question-swapping-x-and-y-axis-for-categorical-predictors-tp4646891.html
Sent from the R help mailing list archive at Nabble.com.

Stephanie,

I'm working on an option for the plot method for Predict that will allow you
to do this.  Note that this approach will not result in very readable
predictor category labels when they are long character strings.  If you are
using linux I can get you a new version of rms with this option as soon as
I've finished it, otherwise I can give you a workaround using source( ).  If
you don't hear from me in a few days please write back.

Frank

stephsus wrote
> Hello all,
> 
> I'm trying to plot the effects of variables estimated by a regression
> model fit individually, and for categorical predictors, the independent
> variable shows up on the y-axis, with the dependent variable on the
> x-axis. Is there a way to prevent this reversal?
> 
> Sample code with dummy data:
> 
> # make dummy data
> set.seed(1)
> x1 <- runif(200)
> x2 <- sample(c(1,2),200, TRUE)
> x3 <- sample(c(0,1),200,T)
> x4 <- runif(200)
> 
> # the dependent variable:
> distance <- (x1/3 + x2 + rnorm(200)^2 - x3 - x4/2)
> 
> # factor two vars, and add to datadist:
> x3 <- factor(x3)
> x2 <- factor(x2)
> 
> d <- datadist(x1,x2,x3,x4)
> options(datadist="d")
> 
> # Make a simple model:
> f <- ols(distance ~ x1 + x2 + x4+ x3, x=T)
> 
> # plot variable effect of a categorical variable:
> plot(Predict(f, x2))
> 
> ^ above step generates a plot with x2 on the y-axis and distance on the
> x-axis, which is the opposite of what I'm aiming for. The continuous
> variables do not have this problem; nor does the plot(Predict(f)) function
> to plot all of the effects at once.
> 
> Thank you so much in advance for your replies! My apologies if this
> question has been answered already; I've tried searching to no avail.
> 
> Best,
> Stephanie
> 
> (Stanford University, Department of Linguistics)




-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context:
http://r.789695.n4.nabble.com/rms-plot-Predict-question-swapping-x-and-y-axis-for-categorical-predictors-tp4646891p4646953.html
Sent from the R help mailing list archive at Nabble.com.