SAS <-> R :: CHEAT SHEET

Introduction New variables, conditional editing Some plotting in R

This guide aims to familiarise SAS users with R. data new_data; new_data <- old_data %>% ggplot( my_data , aes( year , sales ) ) +
R examples make use of tidyverse collection of packages. set old_data; mutate(total_income = wages + benefits) geom_point( ) + geom_line( )
total_income = wages + benefits ;
Install tidyverse: install.packages("tidyverse")
run;
Attach tidyverse packages for use: library(tidyverse)
R data here in ‘data frames’, and occasionally vectors (via c( ) ) data new_data; new_data <- old_data %>%
Other R structures (lists, matrices…) are not explored here. set old_data; mutate(full_time = if_else(hours > 30 , "Y" , "N"))
if hours > 30 then full_time = "Y";
Keyboard shortcuts: <- Alt + - %>% Ctrl + Shift + m else full_time = "N";
run;
Datasets; drop, keep & rename variables data new_data; new_data <- old_data %>%
ggplot( my_data , aes( year , sales ) ) +
geom_point( ) + geom_line( ) + ylim(0, 40) +
set old_data; mutate(weather = case_when( labs(x = "" , y = "Sales per year")
data new_data; new_data <- old_data if temp > 20 then weather = "Warm"; temp > 20 ~ "Warm",
set old_data; else if temp > 10 then weather = "Mild"; temp > 10 ~"Mild",
run; else weather = "Cold"; TRUE ~ "Cold" ) )
run;
data new_data (keep=id); new_data <- old_data %>%
set old_data (drop=job_title) ; select(-job_title) %>%
run; select(id) Counting and Summarising
data new_data (drop= temp: ); new_data <- old_data %>% proc freq data = old_data ; old_data %>%
set old_data; select( -starts_with("temp") ggplot(my_data, aes( year, sales, colour = dept) ) +
table job_type ; count( job_type )
run; For percent, add: geom_point( ) + geom_line( )
C.f. contains( ) , ends_with( ) run;
%>% mutate(percent = n*100/sum(n))
data new_data; new_data <- old_data %>% proc freq data = old_data ; old_data %>%
set old_data; rename(new_name = old_name) table job_type*region ; count( job_type , region )
rename old_name = new_name; run;
run; Note order differs
proc summary data = old_data nway ; new_data <- old_data %>%
Conditional filtering class job_type region ;
output out = new_data ;
group_by( job_type , region ) %>%
summarise( Count = n( ) )
run; ggplot( my_data , aes( year, sales, fill = dept) ) +
data new_data; new_data <- old_data %>% Equivalent without nway not trivially produced geom_col( )
set old_data; filter(Sex == "M") proc summary data = old_data nway ; new_data <- old_data %>%
if Sex = "M"; class job_type region ; group_by( job_type , region ) %>%
run; var salary ; summarise( total_salaries = sum( salary ) ,
output out = new_data Count = n( ) )
data new_data; new_data <- old_data %>% sum( salary ) = total_salaries ;
set old_data; filter(year %in% Lots of summary functions in both languages
run;
if year in (2010,2011,2012); c(2010,2011,2012)) Swap summarise( ) for mutate( ) to add summary data to original data
run;
Combining datasets Note ‘colour’ for lines & points, ‘fill’ for shapes

data new_data; new_data <- old_data %>% ggplot( my_data , aes( year, sales, fill = dept) ) +
set old_data; group_by( id ) %>% geom_col( position = "dodge" ) + coord_flip( )
data new_data ; new_data <- bind_rows( data_1 , data_2 )
by id ; slice(1) set data_1 data_2 ;
if first.id ; run;
run; C.f. rbind( ) which produces error if columns are not identical
Could use slice(n( )) for last
data new_data ; new_data <- left_join( data_1 , data_2 , by = "id")
data new_data; new_data <- old_data %>% merge data_1 (in= in_1) data_2 ;
set old_data; filter(dob > as.Date("1990-04-25")) by id ;
if dob > "25APR1990"d; if in_1 ;
run; run; C.f. full_join( ) , right_join( ) , inner_join( ) C.f. position = "fill" for 100% stacked bars/cols

CC BY SA Brendan O’Dowd • brendanjodowd@gmail.com • Updated 2021-09

Sorting and Row-Wise Operations Dealing with strings
proc sort data=old_data out=new_data; new_data <- old_data %>% data new_data; new_data <- old_data %>%
by id descending income ; arrange( id , desc( income ) ) set old_data; filter( str_detect( job_title , "Health" ))
run; if find( job_title , "Health" );
run;
proc sort data=old_data nodup; old_data <- old_data %>%
by id job_type; arrange( id , job_type)) %>% data new_data; new_data <- old_data %>%
run; distinct( ) set old_data; filter( str_detect( job_title , "^Health" ))
if job_title =: "Health" ;
Note nodup relies on adjacency of duplicate rows, distinct( ) does not
run;
Use ^ for start of string, $ for end of string, e.g. "Health$"
proc sort data=old_data nodupkey; old_data <- old_data %>%
by id ; arrange( id ) %>% data new_data; new_data <- old_data %>%
run; group_by( id ) %>% set old_data; mutate( substring = str_sub( big_string , 3 , 6 ))
slice( 1 ) substring = substr( big_string , 3 , 4 );
run;
Returns characters 3 to 6. Note SAS uses <start>, <length>, R uses <start>, <end>
data new_data; new_data <- old_data %>%
set old_data; group_by( id ) %>% data new_data; new_data <- old_data %>%
by id descending income ; slice(which.max( income )) set old_data; mutate( address = str_replace_all( address , "Street" , "St" ))
C.f.which.min( )
if first.id ; address = tranwrd( address , "Street" , "St" );
Swap to preserve duplicate maxima: … slice.max( income )
run; run;
Alternatively: … filter(income==max(income)) C.f. str_replace( ) for first instance of pattern only

data new_data; new_data <- old_data %>% data new_data; new_data <- old_data %>%
set old_data; mutate( prev_id = lag( id , 1 )) set old_data; mutate( full_name = str_c( first_name , surname , sep = " " ))
prev_id= lag( id ); full_name = catx(" " , first_name , surname );
run; run;
C.f. lead( ) for subsequent rows Drop sep = " " for equivalent to cats( ) in SAS
data new_data; new_data <- old_data %>% data new_data; new_data <- old_data %>%
set old_data; group_by( id ) %>% set old_data; mutate( first_word = word( sentence , 1 ))
by id; mutate( counter = row_number( ) ) first_word = scan( sentence , 1 );
counter +1 ; run;
R example preserves punctuation at the end of words, SAS doesn’t
if first.id then counter = 1;
run; data new_data; new_data <- old_data %>%
set old_data; mutate( house_number = str_extract( address , "\\d*" ))
house_number = compress( address , , "dk" );
Converting and Rounding run;
Wide range of regexps in both languages, this example extracts digits only

data new_data; new_data <- old_data %>%

set old_data ; mutate(num_var = as.numeric("5" )) %>% File operations
num_var = input("5" , 8. ); mutate(text_var = as.character( 5 ))
text_var = put( 5 , 8. ); Operate in ‘Work’ library. Operate in a particular ‘working directory’ (identify using getwd( ) )
run; Use libname to define file locations Move to other locations using setwd( )

data new_data ; new_data <- old_data %>% libname library_name "file_location"; save(data_in_use , file="file_location/saved_data.rda")
set old_data; mutate(nearest_5 = round(x/5)*5) %>% data library_name.saved_data; or
nearest_5 = round( x , 5 ) mutate(two_decimals = round( x , digits = 2) set data_in_use; setwd("file_location")
two_decimals = round( x , 0.01) run; save( data_in_use , file = "saved_data.rda")
run;
libname library_name "file_location"; load("file_location/saved_data.rda" )
Creating functions to modify datasets data data_in_use ;
set library_name.saved_data ;
or
setwd("file_location")
save( ) can store multiple data frames in a
run; load("saved_data.rda")
%macro add_variable(dataset_name); add_variable <- function( dataset_name ){ single .rda file, load( ) will restore all of these
data &dataset_name; dataset_name <- dataset_name %>% proc import datafile = "my_file.csv" my_data <- read_csv("my_file.csv")
set &dataset_name; mutate(new_variable = 1) out = my_data dbms = csv;
new_variable = 1; return( dataset_name ) run;
run; } Both examples assume column headers in csv file
%mend; my_data <- add_variable( my_data )
%add_variable( my_data );
Note SAS can modify within the macro,
whereas R creates a copy within the function
CC BY SA Brendan O’Dowd • brendanjodowd@gmail.com • Updated 2021-09