r-statistics.co by Selva Prabhakaran

About

R for SAS Users: Translate SAS Code to R Step-by-Step

SAS and R solve the same problems — data manipulation, statistical analysis, reporting — but with fundamentally different syntax. This guide translates every common SAS construct into clean R code.

If you've written DATA steps and PROC calls for years, R's syntax will feel foreign at first. But the concepts map directly: a DATA step is a sequence of dplyr verbs, PROC MEANS is aggregate(), and SAS macros become R functions. This guide gives you the translation for every common task.

Why Consider R Alongside SAS?

Factor SAS R
Cost $8,000+/year (Base) Free and open source
Learning curve Familiar if you know it Steeper start, but more flexible
Data step paradigm Row-at-a-time processing Vectorized operations (faster to write)
Visualization SAS/GRAPH, ODS Graphics ggplot2, plotly, lattice (far more flexible)
Packages Fixed SAS modules 20,000+ CRAN packages
Reproducibility Program files Scripts + renv + targets
FDA compliance Industry standard Gaining acceptance (R Validation Hub)
Community Shrinking Growing rapidly

DATA Step → R Data Manipulation

The SAS DATA step is the workhorse for data manipulation. In R, these operations use base R or dplyr.

Creating and Modifying Variables

SAS DATA Step R Equivalent
data new; set old; new <- old
x = a + b; df$x <- df$a + df$b
if age >= 18 then adult = 1; else adult = 0; df$adult <- ifelse(df$age >= 18, 1, 0)
length name $50; Character columns auto-size in R
drop var1 var2; df$var1 <- NULL; df$var2 <- NULL
keep var1 var2; df <- df[, c("var1", "var2")]
rename old=new; names(df)[names(df) == "old"] <- "new"
where age > 18; df <- subset(df, age > 18)
by group; first.group !duplicated(df$group)
by group; last.group !duplicated(df$group, fromLast=TRUE)
retain total 0; total + x; df$total <- cumsum(df$x)
# SAS DATA step equivalent in R df <- data.frame( name = c("Alice", "Bob", "Carol", "Dave", "Eve"), age = c(25, 17, 32, 15, 28), score = c(88, 92, 75, 95, 82) ) # Compute new variables (like SAS assignment statements) df$adult <- ifelse(df$age >= 18, "Yes", "No") df$grade <- ifelse(df$score >= 90, "A", ifelse(df$score >= 80, "B", ifelse(df$score >= 70, "C", "D"))) cat("=== Modified Data ===\n") print(df)


  






Merging / Joining Data







SAS Syntax
R Equivalent






merge a b; by id; (inner)
merge(a, b, by = "id")




merge a(in=x) b(in=y); by id; if x; (left)
merge(a, b, by = "id", all.x = TRUE)




merge a(in=x) b(in=y); by id; if y; (right)
merge(a, b, by = "id", all.y = TRUE)




merge a b; by id; if x and y; (inner)
merge(a, b, by = "id")




set a b; (append/stack)
rbind(a, b)









  


    
# SAS MERGE equivalent
customers <- data.frame(id = 1:4, name = c("Alice","Bob","Carol","Dave"))
orders <- data.frame(id = c(1,2,2,4), amount = c(100, 250, 75, 300))

# Left join (like SAS: merge customers(in=a) orders; by id; if a;)
result <- merge(customers, orders, by = "id", all.x = TRUE)
cat("=== Left Join Result ===\n")
print(result)


    


      

      

    


    


  






PROC Equivalents



PROC MEANS → R







SAS
R






proc means data=df; var x; run;
summary(df$x)




proc means data=df mean std; class group; var x; run;
aggregate(x ~ group, df, function(x) c(mean=mean(x), sd=sd(x)))









  


    
# PROC MEANS equivalent
cat("=== PROC MEANS: mpg by cyl ===\n")
result <- aggregate(mpg ~ cyl, data = mtcars,
  FUN = function(x) c(
    N = length(x),
    Mean = round(mean(x), 2),
    Std = round(sd(x), 2),
    Min = min(x),
    Max = max(x)
  ))
print(result)


    


      

      

    


    


  






PROC FREQ → R







SAS
R






proc freq data=df; tables x; run;
table(df$x)




proc freq data=df; tables x*y / chisq; run;
table(df$x, df$y) + chisq.test()









  


    
# PROC FREQ equivalent
cat("=== One-way Frequency ===\n")
freq <- table(mtcars$cyl)
pct <- round(prop.table(freq) * 100, 1)
print(data.frame(Freq = as.numeric(freq), Pct = as.numeric(pct), row.names = names(freq)))

cat("\n=== Cross-tabulation (cyl * am) ===\n")
ct <- table(Cylinders = mtcars$cyl, Transmission = mtcars$am)
print(ct)
cat("\nChi-square p-value:", round(chisq.test(ct)$p.value, 4), "\n")


    


      

      

    


    


  






PROC REG → R







SAS
R






proc reg data=df; model y = x1 x2; run;
lm(y ~ x1 + x2, data = df)




proc reg; model y = x1 x2 / selection=stepwise;
step(lm(y ~ ., data = df))




proc reg; model y = x1 x2 / vif;
car::vif(model)









  


    
# PROC REG equivalent
cat("=== Linear Regression ===\n")
model <- lm(mpg ~ wt + hp + qsec, data = mtcars)
print(summary(model))


    


      

      

    


    


  






PROC LOGISTIC → R





  


    
# PROC LOGISTIC equivalent
cat("=== Logistic Regression ===\n")
logit <- glm(am ~ mpg + wt + hp, data = mtcars, family = binomial)
print(summary(logit))

cat("\n=== Odds Ratios ===\n")
print(round(exp(coef(logit)), 3))


    


      

      

    


    


  






PROC SORT → R





  


    
# PROC SORT DATA=mtcars; BY cyl descending mpg; RUN;
sorted <- mtcars[order(mtcars$cyl, -mtcars$mpg), ]
cat("=== Sorted by cyl (asc), mpg (desc) ===\n")
print(head(sorted[, c("cyl", "mpg", "hp")], 10))


    


      

      

    


    


  






PROC TRANSPOSE → R





  


    
# PROC TRANSPOSE equivalent
wide_data <- data.frame(
  id = c(1, 1, 2, 2),
  measure = c("height", "weight", "height", "weight"),
  value = c(170, 65, 180, 80)
)

# Wide format (like PROC TRANSPOSE)
wide <- reshape(wide_data, direction = "wide",
                idvar = "id", timevar = "measure", v.names = "value")
cat("=== Wide Format ===\n")
print(wide)


    


      

      

    


    


  






SAS Macros → R Functions



SAS macros generate code dynamically. In R, functions serve the same purpose — but they're simpler and more powerful.







SAS Macro
R Function






%let var = value;
var <- "value"




%macro name(param); ... %mend;
name <- function(param) { ... }




%do i = 1 %to 10;
for (i in 1:10) or lapply(1:10, ...)




%if &cond %then ...;
if (cond) ...




%include "file.sas";
source("file.R")




&var (macro variable)
Regular R variable




%put &var;
cat(var, "\n")









  


    
# SAS macro → R function
# %macro summarize(data=, var=, group=);
#   proc means data=&data; class &group; var &var; run;
# %mend;

summarize <- function(data, var, group) {
  formula <- as.formula(paste(var, "~", group))
  result <- aggregate(formula, data = data,
    FUN = function(x) c(N = length(x), Mean = round(mean(x), 1)))
  print(result)
}

cat("=== Reusable Summary Function ===\n")
summarize(mtcars, "mpg", "cyl")


    


      

      

    


    


  






Complete PROC → R Reference







SAS Procedure
R Equivalent
Package






PROC MEANS
summary(), aggregate()
base




PROC FREQ
table(), prop.table()
base




PROC UNIVARIATE
summary(), shapiro.test()
base




PROC REG
lm()
base




PROC LOGISTIC
glm(family=binomial)
base




PROC GLM
aov(), lm()
base




PROC MIXED
lme4::lmer()
lme4




PROC CORR
cor(), cor.test()
base




PROC SORT
order(), df[order(...), ]
base




PROC TRANSPOSE
reshape(), tidyr::pivot_wider()
base, tidyr




PROC PRINT
print(), head()
base




PROC EXPORT
write.csv()
base




PROC IMPORT
read.csv(), haven::read_sas()
base, haven




PROC SGPLOT
ggplot2::ggplot()
ggplot2




PROC SURVEYSELECT
sample()
base




PROC PHREG
survival::coxph()
survival




PROC FACTOR
factanal()
base




PROC CLUSTER
hclust(), kmeans()
base







FAQ



Can R read SAS .sas7bdat files? Yes. Use haven::read_sas("datafile.sas7bdat"). It preserves variable labels, formats, and missing value types. The sas7bdat package is an alternative but haven is faster and better maintained.



Is R accepted by the FDA for clinical trials? Increasingly, yes. The R Validation Hub (pharmaR.org) provides a framework for validating R packages for regulatory use. Many pharma companies now use R alongside SAS, and some have switched entirely. The FDA accepts R-generated output.



How do I replicate SAS formats in R? SAS formats (like putting labels on numeric codes) are handled by factors in R: df$gender <- factor(df$gender_code, levels = c(1,2), labels = c("Male","Female")). For date formats, use format(date, "%m/%d/%Y").



What's Next




        


        
      


      

        

        
© 2016- Selva Prabhakaran.
          This work is licensed under the Creative Commons License.