Run R from SAS

Run R from SAS

Xin Wei developed a macro called PROC_R which allows R language to be submitted with a SAS program in the base SAS environment. Check out the link for reference. It was designed similar to PROC SQL which lets users to write SQL queries on SAS datasets.

Steps to integrate R with SAS

Step 1 : Download Proc_R Code

Step 2 : Save the file in desired location.

Step 3 : Open the code and update the path of R executable file in the code below.

%macro quit(rpath=%str(C:\Progra~1\R\R-3.3.1\bin\R.exe))

Note : Make sure you use 'Progra~1' instead of 'Program Files' in the path. For example, my R executable file is saved in this path - C:\Program Files\R\R-3.3.1\bin\R.exe but i used this path - C:\Progra~1\R\R-3.3.1\bin\R.exe

Step 4 : Open Base SAS and call proc_R macro. Replace the existing path in the code with your own path where you have saved proc_R code.

%include "C:\Users\Deepanshu\Desktop\proc_R.sas";

Step 5 : Run R inside SAS environment. See the SAS program below -

%Proc_R(SAS2R=,R2SAS=);
cards4;
/********/
/*R Code*/
/********/
;;;;
%quit;

Proc_R Parameters

1. SAS2R   - specifies the names of SAS datasets to be converted to R dataframe. Can be single file name or  multiple files whose names are separated by space.
2. R2SAS   - specifies the names of R data frames to be converted to SAS datasets. Can be single file name or multiple files whose names are separated by space.

Examples : Proc_R

Create a sample data

data example;
input Srl x1-x5;
cards;
1 . 89 85 . 87
2 79 73 74 . .
3 80 82 . 85 .
;
run;

Example 1 : Calculating Summary Statistics

The following program first exports input data in CSV format and R code. Then it imports CSV data from the working directory into R and run the code you put inside %proc_R. Later it exports R log, data frame and image which are later imported in SAS. Here, we are telling SAS to integrate with R to calculate summary statistics on example dataset.

%include "C:\Users\Deepanshu\Desktop\proc_R.sas";

/*Run Summary Statistics*/
%Proc_R(SAS2R=example,R2SAS=);
cards4;
setwd("C:/Users/Deepanshu/Documents")
summary(example)
;;;;
%quit;

proc print data=rlog;
run;

Note : It is required to set the working directory with setwd() function. Update your directory in the setwd() function.

Summary displayed in SAS result window

The output shown in the image above is a truncated one as the real output is large in size.

Example 2 : Data Manipulation with dplyr Package

In the program below, we are loading dplyr library and calculating first non-missing record in each row. The calculation would be done in R and later SAS exports dataframe df.

%Proc_R(SAS2R=example,R2SAS=df);
cards4;
setwd("C:/Users/Deepanshu/Documents")
library(dplyr)
df = mutate(example,nonmiss=coalesce(x1,x2,x3,x4,x5))
;;;;
%quit;

Example3 : Creating Bar Chart

The following code creates a bar chart in R and later displayed in SAS Results Viewer.

%Proc_R(SAS2R=,R2SAS=);
cards4;
setwd("C:/Users/Deepanshu/Documents")
mydata = sample(LETTERS[1:5],16,replace = TRUE)
mydata.count= table(mydata)
barplot(mydata.count)
colors = c("red", "yellow", "green", "violet", "orange", "blue", "pink", "cyan")
barplot(mydata.count, col=colors, main="Main Title ", xlab="X-Axis Title")
;;;;
%quit;

Example 4 : Build ARIMA Model in R

data arimaexample;
input sales @@;
cards;
360358357
374371385
385389398
400412424
418412408
420424438
435438446
451456470
457448440
456457469
473477483
490498503
;
run;

%Proc_R(SAS2R=arimaexample,R2SAS=Forecastmodel);
cards4;
setwd("C:/Users/Deepanshu/Documents")
library(forecast)

# Convert it to ts object
dat = data.frame(arimaexample)
tsdata = ts(dat)

# Plot time series data
plot.ts(tsdata)
tsdisplay(tsdata)

# Box lambda Transformation
lambda = BoxCox.lambda(tsdata)

#lambda close to 1 means no transformation required
tsdata2 = BoxCox(tsdata, lambda=lambda)
plot.ts(tsdata2)

# Unit Ratio Tests
library(tseries)
adf = adf.test(tsdata2)
kpss = kpss.test(tsdata2)

# Number of Difference Required to make data stationary
ndiffs(tsdata2)
tsdata3 = diff(tsdata2, differences = 1)
final = auto.arima(tsdata, trace= TRUE, ic ="bic", approximation = FALSE, stepwise = FALSE)

# predict the next 5 periods
Forecastmodel = forecast.Arima(final, h = 5)

;;;;
%quit;

proc print data= rlog;
run;

Example 5 : Supports Multiple datasets / dataframes

In the macro, we can also specify multiple SAS datasets or data frames. In the code below, we have specified multiple SAS datasets that would be imported into R and then perform some calculation and later multiple R data frames would be exported.

%Proc_R(SAS2R=example arimaexample,R2SAS= df1 df2);
cards4;
setwd("C:/Users/Deepanshu/Documents")
df1=data.frame(mean(as.numeric(arimaexample$sales)))
df2=data.frame(sum(is.na(example)))
;;;;
%quit;
proc print data = df1;
proc print data = df2;
run;