Transcript
Page 1: Moving From SAS to R Webinar Presentation - 07Aug14

Moving from SAS to R

Thomas W. Dinsmore

Seth Mottaghinejad

August 7

Page 2: Moving From SAS to R Webinar Presentation - 07Aug14

Poll #1 What analytics platforms are you currently using?

(please check all that apply)

• R/RRE

• SAS

• SPSS

• Tibco/Spotfire

• Other

2

Page 3: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential Moving from SAS to R

Introduction

Why choose R over SAS?

Concerns about R

How we can help

How to Migrate from SAS to R

Customer Success

Questions

Page 4: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential R: Explosive Growth

Source: Rexer Data Miner Survey, 2007-2013

Use R

R is Primary Tool

70%

Page 5: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential Who uses R for analytics today?

5

Page 6: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

6

R Penetrates Enterprise Analytics

0

200

400

600

800

1,000

1,200

1,400

1,600

2014 2015 2016 2017 2018

Top 10 Global Bank – Planned Analytic Seats

SAS R

Page 7: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

7

Why choose R over

SAS?

Page 8: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

8

Your new analysts already know R.

Page 9: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

9

R is a comprehensive

analytics platform.

Page 10: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

10

R is flexible and customizable.

Page 11: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

11

R is free; SAS is expensive.

Page 12: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

12

Concerns about R

Page 13: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

13

“We don’t know what’s in the code.”

Page 14: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

14

“We’ll be on our own.”

Page 15: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

15

“R doesn’t scale.”

Page 16: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

16

“We have a lot of SAS programs.”

Page 17: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

17

How we can help:

“We don’t know what’s in the code.”

“We’ll be on our own.”

“R doesn’t scale.”

“We have a lot of SAS programs.”

Build Assurance

Technical Support

Distributed Platform

Migration Services

Page 18: Moving From SAS to R Webinar Presentation - 07Aug14

Are you currently exploring alternative analytics

platforms?

• Yes

• No

Poll #2

Page 19: Moving From SAS to R Webinar Presentation - 07Aug14

Migrating from SAS to R: What you need to know

19

Page 20: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential About me

20

Stats background

Over 5 years R and SAS

experience

Worked on migrating SAS

projects to R

Teach a class on R for SAS

users

Passionate advocate

bringing R to industry

Page 21: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

21

You may be… • New to R?

• Using SAS and • think it is too expensive

• think it does not integrate well with other apps

• have a hard time attracting younger talent

• curious about R

• thinking of migrating to R

• in the process of migrating to R

• want to scale up to a Hadoop cluster or a traditional grid

• Migration from SAS to R is doable.

Page 22: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

22

SAS vs R and RRE • Conceptually, SAS and R are very different

languages, yet you can leverage a lot of what you know. Revolution Analytics has training and services to help you learn to work in R.

• Often when switching from SAS to R we give up a pre-packaged solution in return for a flexible, robust solution that you can easily augment.

• Writing code that works in R is easy, but writing production-level code in R is more difficult.

• Through its ScaleR package, RRE brings to the table many of the features that make R an enterprise solution.

Page 23: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

23

SAS vs R and RRE • The explicit logic is the logic for implementing the

project independent of any specific code. The implicit logic is the logic for implementing the project based on reading and understanding the code.

• Can you diagram your project’s workflow (explicit logic)?

• How much do you value adherence to the implicit logic (SAS code)?

• If you had to rewrite the project, would you write the code in the same way or is there room for simplifying?

• Was the project the work of a single person or did multiple people work on it?

Page 24: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

24

Migrating from SAS: DATA Step • Working with data (create, transform, and

reshape) is straightforward in R.

• If you use PROC SQL for many of your data operations, you may be able to bring most of it into R without having to rewrite it in R code, although this may not be the most efficient way.

• You must act with caution in some cases, such as missing data or formatting issues.

• How much data-processing do you perform?

• What sorts of data sets do you import/read from?

Page 25: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

25

Migrating from SAS: PROCs • Except for a few notable cases (such as PROC

IMPORT/EXPORT, TRANSPOSE) most PROCs in SAS are used to summarize data or to analyze data (modeling and data-mining).

• Some statistical procedures may be implemented in a slightly different way in SAS vs R, resulting in tiny differences in computations. If such differences matter, we can always build on an existing solution or write our own. Creating a custom-made solution is far more easier in R than in SAS.

• What statistical procedures do you use?

• Do you want to exactly replicate your SAS code?

Page 26: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

26

Migrating from SAS: Macros • Macros in SAS are a “separate” language, built on top

of the SAS language. You write SAS code to manipulate and analyze data, and you write macros to manipulate the SAS code itself, with the intention of automating the code.

• There is no “R macro language”: R is flexible enough to also be its own macro language. However, in the case of more advanced macros, this often means that the converted R code will look very different from its SAS counterpart.

• Do you use macros? If so, can you provide examples?

Page 27: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

27

Migrating from SAS: ODS • ODS is SAS’s way of exporting results to other

formats. It can be used as part of a pipeline or for reporting.

• The fact that R is open-source makes it even easier to create hookups to external formats (Excel, PDFs, HTML, etc.), which results in even more flexibility.

• Can you describe your pipeline/workflow?

• What formats do you export to? Who is the target audience (e.g. internal vs external)?

Page 28: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

28

Migrating from SAS: What else? • SAS has more components than covered so far, e.g.

SAS Enterprise Miner, SAS EG, SAS IML, add-ons for design and analysis of experiments, and more.

• R’s ease of integration with third party tools makes it very easy to extend its capability.

• What are some specific SAS features that you want to migrate into R?

• What are some concerns you might have about migrating?

• How are you preparing for the migration?

Page 29: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

29

Handover of the R project • For a successful handover, we need to think of ways to test

the R code to ensure that it can replicate the SAS code.

• Documentation will be provided and some on-site knowledge transfer may be needed.

• Training is also an option.

• How much of your data will you share?

• Is the data you will share representative of your “best data” or your “worst data”?

• What are your success criteria?

• Who will be in charge of testing the code?

• How do you envisage the knowledge transfer?

Page 30: Moving From SAS to R Webinar Presentation - 07Aug14

In what scenario would you consider supplementing your

production analytics platform? (all that apply)

• New data environment (Hadoop)

• Research / Data Exploration

• Improve performance for production systems

• Lower costs for production systems

• New production application development

Poll #3

Page 31: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

31

Customer Success

Page 32: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

32

Old Solution New Solution

Page 33: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

33

Old Solution New Solution

Page 34: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

34

Old Solution New Solution

Page 35: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential Moving from SAS to R

Introduction

Why choose R over SAS?

Concerns about R

How we can help

How to Migrate from SAS to R

Customer Success

Questions

Page 36: Moving From SAS to R Webinar Presentation - 07Aug14

Revolution Confidential

36

Page 37: Moving From SAS to R Webinar Presentation - 07Aug14

Join us in September Merkle will share their experience building a multi-channel

customer analytics system with RRE, Cloudera and

Tableau

You can register on the Revolution Analytics website next week.

23SEP14 at 10:00 AM, Pacific

37


Top Related