Permalink
Switch branches/tags
Nothing to show
Find file
8fab97a Feb 17, 2017
30 lines (15 sloc) 4.75 KB

R Onboarding for SAS Users

First, I'd like to welcome another SAS user to the R Programming course in the Johns Hopkins Data Science Specialization on coursera.org. You are following in the footsteps of a long line of (mostly) university-trained SAS users into the wild world of R. Your journey will both be fun and frustrating, as you discover that R is much different than the record-oriented processing that is the base of the SAS DATA step, based on its heritage of reading raw data files based on IBM mainframe file structures of the 1970s. On the fun side, you'll be amazed at the power of R given its cost (free), and the wide variety of packages that have been contributed by the R community.

As to my background in SAS, I started statistical programming back when I was an undergraduate, using SAS 82 (SAS releases were named by year before the Multi-Vendor Architecture was implemented in 1985). My first work in SAS was as an undergraduate Research Assistant on a National Science Foundation project to study political contributions to Illinois legislative candidates that supported The Party Network: The Robust Organization of Illinois Republicans, by Mildred A. Schwartz.

Since then I've written a wide variety of software in SAS, including a principal components pre-processor to an expert system that discovered a secret price test on various laundry detergent products that was being conducted by a large consumer package goods manufacturer. I have also taught Base SAS and SAS Macro Language in corporate environments.

If you've ever used SAS/IML, the interactive matrix language, you'll find many parts of R are like SAS/IML. Since SAS/IML uses a programming model that is significantly different from Base SAS, as you learn R you will experience a learning curve that is similar to a SAS user learning to work with SAS/IML.

The bottom line is that thinking in R is significantly different than thinking in SAS. To highlight some of the differences, take a look at Thinking in R versus Thinking in SAS, and Why is R Harder than SAS?

Before you begin working on the first programming assignment, read the following, in order:

Strategy for the Programming Assignments,

Derek Franks' Tutorial for Those Struggling with Assignment 1, and

Breaking Down Pollutantmean(), which provides a detailed walkthrough illustrating how to organize this function.

You might also enjoy Commercial Statistics Packages: An Historical Perspective, but it's not required to get through the assignment.

Another interesting but not required read is Linda Burtch's article SAS vs. R vs. Python: Which Tool do Analytics Pros Prefer? on the kdnuggets.com website.

Finally, if you're interested in understanding how R compares to other statistics packages for a variety of statistical analyses, take a look at the Data Analysis Examples page on UCLA's Institute for Digital Research and Education website.

After completing the first programming assignment, you'll discover that the R-based solution to pollutantmean() is significantly less complicated than a SAS-based solution. How can I say this? A SAS Version of pollutantmean()? illustrates the code needed to replicate pollutantmean() in SAS. Believe it or not, one can complete the pollutantmean() assignment in a single R statement. It's a complex statement of nested functions, but it evaluates as a single R statement.

The SAS version of pollutantmean() is a lot more complicated than the most elegant R solution, in part because the design of SAS forces one into using SAS macros because SAS does not make it easy for end users to write their own SAS procedures. While it is technically possible for an end user to write a SAS procedure (see Betancourt 1985), it requires a significant effort.