topic Re: Translate code from R to SAS in Statistical Procedures

Translate code from R to SAS

desireatem — Thu, 11 Sep 2014 16:01:07 GMT

How do I translate this R code to SAS:

library(lme4)

sim<-function(m,n,b0=1,b1=-.5,b2=-1,testlow=1,testhi=10,xtest,xolftrue){

a0hat<-0

b0hat<-0

b1hat<-0

b2hat<-0

d0hat<-0

c0hat<-0

c1hat<-0

c2hat<-0

c3hat<-0

for(i in 1:m){

olf<-runif(n,1,10)

olftrue<-as.vector(t(matrix(olf,nrow=n,ncol=10,byrow=F)))

test<-sample(testlow:testhi,n,replace=T)

olftest<-ifelse(olf<test,test,0)

olfind<-ifelse(olf<test,1,0)

olft<-matrix(0,nrow=n,ncol=10,byrow=T)

olft[cbind(1:n,olftest)]<-1

olft<-t(apply(olft,1,cumsum))

olft<-as.vector(t(matrix(olft,nrow=n*10,ncol=1,byrow=T)))

x<-rnorm(n,3,1)

mx<-matrix(x,nrow=n*10,ncol=1,byrow=T)

x<-as.vector(mx)

alpha<-rnorm(n,0,.5)

malpha<-matrix(alpha,nrow=n*10,ncol=1,byrow=T)

alpha<-as.vector(malpha)

eps<-rnorm(n*10,0,1)

time<-matrix(1:10,nrow=n*10,ncol=1,byrow=T)

time<-as.vector(time)

time1<-pmin(time,olftrue)

time2<-pmax(time-olftrue,0)

y<-alpha+b0*x+b1*time1+b2*time2+eps

id<-as.vector(t(matrix(1:n,nrow=n,ncol=10,byrow=F)))

ot<-olft*time

try<-lmer(y~1+(1|id)+x+time1+time2)

a0hat<-a0hat+fixef(try)[[1]]

b0hat<-b0hat+fixef(try)[[2]]

b1hat<-b1hat+fixef(try)[[3]]

b2hat<-b2hat+fixef(try)[[4]]

timenew<-time-as.vector(t(matrix(olftest,nrow=n,ncol=10,byrow=F)))

olfind<-as.vector(t(matrix(olfind,nrow=n,ncol=10,byrow=F)))

try2<-lmer(y~1+(1|id)+x+timenew+olfind+timenew*olfind)

d0hat<-d0hat+fixef(try2)[[1]]

c0hat<-c0hat+fixef(try2)[[2]]

c1hat<-c1hat+fixef(try2)[[3]]

c2hat<-c2hat+fixef(try2)[[4]]

c3hat<-c3hat+fixef(try2)[[5]]

}

Re: Translate code from R to SAS

LinusH — Thu, 11 Sep 2014 16:18:08 GMT

You probably have to learn IML programming, or hire one to do it for. There are some tools out there, but I doubt none can convert all R program to SAS out of the box.

Re: Translate code from R to SAS

lvm — Thu, 11 Sep 2014 16:46:44 GMT

There is no automated way. I doubt if any of our readers would want to spend a bunch of time doing this. You would first have to learn PROC IML and do it. There is another possibility: You can call R from within PROC IML. See the User's Guide, or check this out:

SAS/IML(R) 9.22 User's Guide

It is easy to embed R within an IML run, get the results and use with SAS.

Re: Translate code from R to SAS

Rick_SAS — Thu, 11 Sep 2014 17:14:16 GMT

Did the program editor eat some of your code? It looks like you are missing some closing brackets (the function and the for loop).

It looks like this is a simulation of a regression model. Presumably a mixed model, since it uses the lme4 package. Part of the missing code probably estimate parameters to obtain the sampling distribution.

You need to say how sophisticated your SAS programming skills are, and if you know SAS/IML or only the DATA step. It would also be nice if you tell us what you ultimately hope to accomplish.

There are two chapters about simulating regression models in the book Simulating Data with SAS, as well as many tips for efficient simulation There are also examples at this blog: Sampling and Simulation - The DO Loop Here is a blog post that is similar to your example, but it is for a logistic model: Simulate many samples from a logistic regression model - The DO Loop

Re: Translate code from R to SAS

Rick_SAS — Thu, 11 Sep 2014 17:15:29 GMT

Here is a more recent version of the documenation: SAS/IML(R) 13.1 User's Guide