Skip to content

How to pass a variable to a function which use variable name as a parameter

4 messages · wong jane, John McKown, William Dunlap +1 more

#
There are functions which use variable names as parameters in some R
packages. However, if the variable name is stored in another variable, how
can I pass this variable to the function. Taking the "rms" package as an
example:

library(rms)
n <- 1000
age <- rnorm(n, 50, 10)
sex <- factor(sample(c('female','male'), n,TRUE))

y <- rnorm(n, 200, 25)
ddist <- datadist(age, sex)
options(datadist='ddist')
fit <- lrm(y ~ age)
Predict(fit, age, np=4)
options(datadist=NULL)

Here "age" was a variable name passed to Predict() function, but if "age"
was stored in variable "var", that is, var <- "age", how can I pass "var"
to Predict() function? The purpose is that I want to change the parameter
of Predict()  in a loop.
#
On Tue, May 26, 2015 at 5:14 AM, wong jane <jane.wong083 at gmail.com> wrote:

            
?Please turn off HMTL email. The forum software doesn't really like it.

What you want is the "get()" function.


var<-"age"

?
?
library(rms)
n <- 1000
age <- rnorm(n, 50, 10)
sex <- factor(sample(c('female','male'), n,TRUE))

y <- rnorm(n, 200, 25)
?#?
ddist <- datadist(age, sex)
?ddist <- datadist(get(var),sex)?
options(datadist='ddist')
fit <- lrm(y ~ age)
?#?
Predict(fit, age, np=4)
?Predict(fit,get(var),np=4)?
options(datadist=NULL)
#
One way to use variable names in functions like Predict() that
do not evaluate their arguments in the standard way is to use
do.call() along with as.name().  E.g.,
  varName<-"age"
  do.call("Predict", list(fit, as.name(varName), np=4))})
gives the same result as
  Predict(fit, age, np=4)



Bill Dunlap
TIBCO Software
wdunlap tibco.com
On Tue, May 26, 2015 at 3:14 AM, wong jane <jane.wong083 at gmail.com> wrote:

            

  
  
#
On May 26, 2015, at 8:30 AM, William Dunlap wrote:

            
Here's another approach, developed after looking at the code and the help page, where I noticed that the factors argument was given an an alternate route for specifying the variables for evaluation. The requirements for the 'factors' argument are that it be provided as a named list and have either specific values or NA.

 var="age"
 vlist <- list(dum=NA)
 names(vlist)=var
 
 Predict(fit, factors=vlist, np=4)
#----- same result -------
       age     yhat    lower    upper
1 25.04901 6.834836 4.856497 8.813174
2 41.14259 6.882369 4.919433 8.845306
3 57.23618 6.929903 4.967118 8.892688
4 73.32977 6.977437 4.999548 8.955325

Response variable (y): log odds 

Limits are 0.95 confidence limits

-- 
David.