THE ISSUES ADDRESSED

The common misconceptions we come across at our academic institutions left us wanting to address the pros and cons of both languages. There could very well be even more issues at play, but the ones we chose to focus on seem to be the most common. We hope to dispel these and possibly provide new information to those who have not kept up with R or SAS.

NEW ADVANCES IN STATISTICAL METHODOLOGY

SAS

PROs: Software and algorithms are thoroughly tested and SAS has tech support to address user needs quickly. SAS tries to implement new methodology into existing procedures when it makes sense to do so, as an option or additional statement, so that users do not need to learn yet another

PROC. SAS also releases newsletters detailing new advances in the software.

CONs: The delay in new advances until the next release of the software.

R

PROs: Users can implement new methodology quickly or find a package from someone who already has. Easy to teach and understand the new methodology because students can see the functions right there in the code.

CONs: The documentation of the new advances in R is user-based, so new methodology is not well organized and often has not been thoroughly tested. The developers are spread out rather than working together locally.

For this case the cons of one are really the pros for the other. In the case of R some would argue that they can see exactly what is going on in the code, and this is true if the user has the background to understand it. However for SAS while the PROCs are pre-packaged, the documentation has an enormous amount of detail about the mathematics involved for every single statement and option. If a user really wanted to get under the hood, the detail is all there and very easy to access in the technical documentation. A student or user blindly running code should be a concern in both languages. Someone who would run a PROC without understanding what they were doing would most likely run a pre-written R package without paying attention to the functions it was calling.

GRAPHICS

SAS

PROs: The graphics in SAS are becoming increasingly flexible, polished and easy to use. In some PROCs, turning on ODS Graphics will automatically produce graphics without requiring additional code. This gives users a choice of using the defaults graphs or creating their own.

CONs: The template language behind the graphics can be overwhelming and somewhat difficult to use, especially for beginners. New advancements, such as interactive graphs, can also still be difficult for beginners.

R

PROs: Easy to create polished graphics. Graphs can also be run through loops to create animations.

CONs: R graphics are not attached to statistical analyses, so graphs and analyses take place separately. The user must decide which graph is most appropriate, which can be beneficial or not depending on the user’s background in statistics. This is more subjective to the user’s preference, though modifying the graph to achieve specific dimensions or aspects is not necessarily an easy task.

Prior to version 9.2 of SAS the graphics were one of the main reasons that people gravitated toward R. One of the best features of R has been, and still is, the quality and ease of its graphics. However the current state of SAS/GRAPH with ODS graphics and SG procedures are have revived the graphics capabilities of the software.

Using ODS graphics with the PROCs allows users to easily generate graphics that relate to the analysis that they are performing. New PROCs for creating more specific graphs, such as PROC SGPLOT, SGPANEL, and SGSCATTER are quickly growing; however they require a reasonable amount of coding. Additionally there are other great graphics options in SAS such as SGDESIGNER and SAS Enterprise Guide.

FUNCTIONS AND REUSABLE CODE

SAS

PROs: SAS has an extensive function library as well as the ability to write user based functions that can be called in DATA and PROC steps. Also the macro programming language with massive scope and capability in which variables can be global or local.

CONs: Writing user based functions and detailed macros requires a fairly extensive amount of programming knowledge to ensure accuracy.

R

PROs: In R the ability to write any function is easy. Users can share functions with others by submitting to R-CRAN.

CONs: Writing user based functions requires a detailed amount of knowledge to ensure accuracy. Variables are strictly local in scope.

In this case the two languages have virtually the same pro and con. In the past SAS users who wanted to run their own functions relied heavily on the macro programming language to get the job done. This was seen as ineffective and clunky by R users. However as of SAS version 9 PROC FCMP allows users to write their own functions, and as of version 9.2 these functions can be called in DATA steps as well as PROCs. This is useful for simple statistical functions, and also more complicated statistical functions which can be implemented using the IML language. Both languages face the issue of being able to carry functions out correctly and in an efficient manner. This requires detailed knowledge of the process of which the user is trying to capture in the function. This is good from the view that a programmer needs to know what they are programming, yet dangerous in the sense that others can download a SAS macro or R package and use it while having no idea how it works or if it is working correctly for them.

However with proper understanding the sharing of functions and macros makes extending them to particular needs very handy.

FREEWARE

SAS

PROs: SAS has an OnDemand version of its software that is free to degree granting institutions.

CONs: SAS proper and JMP are not free. OnDemand has limitations on which operating system it will run, and has been reported to be slow at times.

R

o PROs: R is free.

o CONs: Open source software can be a security concern for large companies.

The free alternative offered by SAS to universities is promising in terms of keeping it a viable option for professors to use in the classroom. The installation process and speed of OnDemand needs attention. That said, SAS and JMP are still not free and a company would still need to license the software. R can be installed for free. Many bloggers covering the debate have pointed out that to convert an existing company who uses SAS to R would be a waste of resources and money that far exceed annual licensing costs. The re-writing of code, transitioning the team, hiring new experts, etc…Also companies that have strict analytical vetting requirements would be better suited toward SAS.

Small companies with no existing analytical infrastructure could take a hard look at whether to invest in licensed software with extensive history and resources, or go with free software that would require somewhat of an investment in advanced knowledge of the staff to create, manipulate and run the coding. In the end, time, money and these two options may very well break even.

USER SUPPORT

SAS

PROs: SAS has extensive online documentation, expert technical support, professional training courses, many excellent books in press, and a tight knit user group and web based community. Problems can be addressed to SAS directly via tech support who replies very quickly and will work with the user to solve the problem.

CONs: We really couldn’t think of any.

R

PROs: R has good books by example, online user documentation, R mailing list, and R meet-ups.

CONs: Users rely on what others put out there about the software. There is a disconnect in the

world-wide user group because the developers are so spread out. Packages are not written by the

R Development Core-Team therefore they are not well polished and some could have questionable validity. It is also difficult to direct an issue to a particular person or support system.

The excellent support provided by SAS highlights its customer oriented design. The strength of SAS’s support makes it ideal for beginners, and its extensive detail is useful for advanced users. R’s disorganized documentation and lack of technical support make finding help a challenge. This is a big trade-off for the developer oriented design of R.

DATA MANIPULATION

SAS

PROs: SAS can handle any data set or data type. DATA steps are designed purely for data management and therefore SAS excels at it. SAS handles big datasets well, with a variety of options for working with them; merging and PROC SQL have a quick run times.

CONs: Coding can be a change of thinking in that SAS has an underlying loop for the DATA step.

R

PROs: R is starting to be more capable of working with big data. It is designed well for matrix manipulation and has fast sorting times. R also excels at simulation based analysis.

CONs: The design of R was focused around statistical computing and graphics, so data management tends to be time consuming and not as clean as SAS. One of the main reasons for this is the large number of different types of data structures which can make data manipulation in R a more difficult concept to grasp.

Somehow data management has been overlooked as an essential aspect of statistical programming, yet it is extremely important since real life data is very rarely clean and ready for analysis. Students who have used solely R have an unrealistic expectation of the state of the data they receive. Learning SAS is an excellent way to learn how to combat more realistic data sets. SAS can manage and analyze big and messy data, whereas R is mostly centered around analysis.

The object oriented data structure of R can prove problematic when dealing with messy data and it does not have an internal looping structure. Merging complicated datasets with large amounts of missing data then creating and modifying variables is a common use of SAS that utilizes standard tools. However, doing complex data manipulation in R is not standard and can turn into a nasty process.

Which software has better run time between SAS and R seems to depend on the task. Options such as MEMLIB can be set in SAS to use main memory as R does, rather than the hard drive to improve run time. R does not have thishard drive memory option and must use main memory.

INSTALLATION

SAS

PROs: SAS comes as a packaged install with the components that are specified according to analytic needs and licensing. The updating of the license is very easy.

CONs: The initial installation of SAS or updating to a new version of the software can be long and difficult. Magnify this by 1,000 in the context of explaining this process to a group of students new to the software who just want to run it on their laptops for class. Proper SAS also does not run on Macs, at least not easily, which are becoming an increasingly prominent laptop choice in the classroom.

R

PROs: R is easy to install and runs on Windows, Mac and Unix. R installs very quickly. This is also true for RStudio, a common and useful user interface for R that many find preferable to R.

CONs: You have to know about the packages that will meet your needs, seek them out, install them, and then research what they can do. As of this writing there are 4,379 packages available (and growing every day) which, while providing many options, can be difficult and time consuming when searching among all of them.

SAS can be difficult for users to obtain and the initial installation is sometimes tricky. However once it is installed problems with the software are rare and there are no additional packages or steps for specific analyses—those are already installed. R is the opposite, it is very easy to install, but packages are required for additional analyses and the time saved from installation can potentially be lost up in this area.

REPORTING

SAS

PROs: SAS provides many useful procedures for creating detailed and polished reports.

CONs: Some of the more detailed reporting procedures, such as TABULATE and REPORT, have a learning curve that takes place before being able to utilize them correctly.

R

PROs: R includes Sweave which makes it possible to create a PDF containing text, tables, and charts by including LaTeX markup and R commands. Another new package knitr allows the quick creation of web content with less formatting issues

CONs: R does not have a defined way of producing reports, and to do so would involve a serious programming investment. Reports seem to be a relatively new idea in R so are not yet as easy or

quick to implement as in SAS. Sweave and knitr are leaders in this area for R, yet they can prove difficult to learn.

Users whose tasks revolve heavily around creating reports should consider this difference. While some time must be invested to learn SAS reporting procedures, once mastered they prove invaluable and highly flexible. The R programming that would be required to create reports from the ground up may not outweigh the time investment in getting to know the SAS report procedures.

We see the solution to the R vs. SAS debate as three-fold. First, we need to understand as a statistical programming community that there is no clear winner. Both packages have their strengths and weaknesses. They need to co-exist and we in academia also need to teach them as they co-exist. Students will be better served in their degrees if they can explore their options and be able to use their skills as appropriate. By only teaching one methodology we are limiting them and at the very least making it more difficult for them to live up to their potential. Second, users need to keep their technology toolkit up to date.

 

 

 

---

Data Science Simplified: Principles and Process

In 2006, Clive Humbly, UK Mathematician, and architect of Tesco’s Clubcard coined the phrase “Data is the new oil. He said the following:

”Data is the new oil. It’s valuable, but if unrefined it cannot be used. It has to be changed into gas, plastic, chemicals, etc. to create a valuable entity that drives profitable activity; so, must data be broken down, analyzed for it to have value.”

The iPhone revolution, growth of the mobile economy, advancements in Big Data technology has created a perfect storm. In 2012, HBR published an article that put Data Scientists on the radar. 

The article Data Scientist: The Sexiest Job of the 21st Century labeled this “new breed” of people; a hybrid of data hacker, analyst, communicator, and trusted adviser.
Every organization is now making attempts to be more data-driven. Machine learning techniques have helped them in this endeavor. I realize that a lot of the material out there is too technical and difficult to understand. In this series of articles, my aim is to simplify Data Science. I will take a cue from the Stanford course/book (An Introduction to Statistical Learning). This attempt is to make Data Science easy to understand for everyone.

In this article, I will begin by covering fundamental principles, general process and types of problems in Data Science.

Data Science is a multi-disciplinary field. It is the intersection between the following domains:

• Business Knowledge
• Statistical Learning aka Machine Learning
• Computer Programming

The focus of this series will be to simplify the Machine Learning aspect of Data Science. In this article, I will begin by covering principles, general process and types of problems in Data Science.

Key Principles:

 

 

• Data is a strategic asset: This concept is an organizational mindset. The question to ask is: “Are we using the all the data asset that we are collecting and storing? Are we able to extract meaningful insights from them?”. I’m sure that the answers to these question are “No”. Companies that are cloud born are intrinsically data-driven. It is in their psyche to treat data as a strategic asset. This mindset is not valid for most of the organization.
• A systematic process for knowledge extraction: A methodical process needs to be in place for extracting insights from data. This process should have clear and distinct stages with clear deliverables. The Cross Industry Standard Process for Data Mining (CRISP-DM) is one such process.
• Sleeping with the data: Organisations need to invest in people who are passionate about data. Transforming data into insight is not alchemy. There are no alchemists. They need evangelists who understand the value of data. They need evangelists who are data literate and creative.They need folks who can connect data, technology, and business.
• Embracing uncertainty: Data Science is not a silver bullet. It is not a crystal ball. Like reports and KPIs, it is a decision enabler. Data Science is a tool and not a means to end. It is not in the realm of absolute. It is in the realm of probabilities. Managers and decision makers need to embrace this fact. They need to embrace quantified uncertainty in their decision-making process. Such uncertainty can only be entrenched if the organizational culture adopts a fail fast-learn fast approach. It will only thrive if organizations choose a culture of experimentation.
• The BAB principle: I perceive this as the most important principle. The focus of a lot of Data Science literature is on models and algorithms. The equation is devoid of business context. Business-Analytics-Business (BAB) is the principle that emphasizes the business part of the equation. Putting them in a business context is pivotal. Define the business problem. Use analytics to solve it. Integrate the output into the business process. BAB.

Process:

 

 

 

Taking a cue from principle #2, let me now emphasize on the process part of data science. Following are the stages of a typical data science project:

1. Define Business Problem

Albert Einstein once quoted “Everything should be made as simple as possible, but not simpler”. This quote is the crux of defining the business problem. Problem statements need to be developed and framed. Clear success criteria need to be established. In my experience, business teams are too busy with their operational tasks at hand. It doesn’t mean that they don’t have challenges that need to be addressed. Brainstorming sessions, workshops, and interviews can help to uncover these challenges and develop hypotheses. Let me illustrate this with an example. Let us assume that a telco company has seen a decline in their year-on-year revenue due to a reduction in their customer base. In this scenario, the business problem may be defined as:

• The company need grow the customer base by targeting new segments and reducing customer churn.

2. Decompose To Machine Learning Tasks

The business problem, once defined, needs to be decomposed to machine learning tasks. Let’s elaborate on the example that we have set above. If the organization needs to grow our the customer base by targeting new segments and reducing customer churn, how can we decompose it into machine learning problems? Following is an example of decomposition:

• Reduce the customer churn by x %.
• Identify new customer segments for targeted marketing.

3. Data Preparation

Once we have defined the business problem and decomposed into machine learning problems, we need to dive deeper into the data. Data understanding should be explicit to the problem at hand. It should help us with to develop right kind of strategies for analysis. Key things to note is the source of data, quality of data, data bias, etc.

4. Exploratory Data Analysis

A cosmonaut traverses through the unknowns of the cosmos. Similarly, a data scientist traverses through the unknowns of the patterns in the data, peeks into the intrigues of its characteristics and formulates the unexplored. Exploratory data analysis (EDA) is an exciting task. We get to understand the data better, investigate the nuances, discover hidden patterns, develop new features and formulate modeling strategies.

5. Modelling

After EDA, we move on to the modeling phase. Here, based on our specific machine learning problems, we apply useful algorithms like regressions, decision trees, random forests, etc.

6. Deployment and Evaluation

Finally, the developed models are deployed. They are continuously monitored to observe how they behaved in the real world and calibrated accordingly.

Typically, the modeling and deployment part is only 20% of the work. 80% of the work is getting your hands dirty with data, exploring the data and understanding it.

Machine Learning Problem Types

 

 

 

In general, machine learning has two kinds of tasks:

Supervised Learning

Supervised learning is a type of machine learning task where there is a defined target. Conceptually, a modeler will supervise the machine learning model to achieve a particular goal. Supervised Learning can be further classified into two types:

• Regression: Regression is the workhorse of machine learning tasks. They are used to estimate or predict a numerical variable. Few examples of regression models can be:

• • What is the estimate of the potential revenue next quarter?
  • How many deals can I close next year?
• Classification: As the name suggests, classification models classify something. It is estimated which bucket something is best suited. Classification models are frequently used in all types of applications. Few examples of classification models are:

• Spam filtering is a popular implementation of a classification model. Here every incoming e-mail is classified as spam or not spam based on certain characteristics.
• Churn prediction is another important application of classification models. Churn models used widely in telcos to classify whether a given customer will churn (i.e. cease to use the service) or not.

Unsupervised Learning

Unsupervised learning is a class of machine learning task where there are no targets. Since unsupervised learning doesn’t have any specified target, the result that they churn out may be sometimes difficult to interpret. There are a lot of types of unsupervised learning tasks. The key ones are:

• Clustering: Clustering is a process of group similar things together. Customer segmentation uses clustering methods.
• Association: Association is a method of finding products that are frequently matched with each other. Market Basket analysis in retail uses association method to bundle products together.
• Link Prediction: Link prediction is used to find the connection between data items. Recommendation engines employed by Facebook, Amazon and Netflix heavily use link prediction algorithms to recommend us friends, items to purchase and movies respectively.
• Data Reduction: Data reduction methods are used to simplify data set from a lot of features to a few features. It takes a large data set with many attributes and finds ways to express them in terms of fewer attributes.

Machine Learning Task to Models to Algorithm

Once we have broken down business problems into machine learning tasks, one or many algorithms can solve a given machine learning task. Typically, the model is trained on multiple algorithms. The algorithm or set of algorithms that provide the best result is chosen for deployment.

Azure Machine Learning has more than 30 pre-built algorithms that can be used for training machine learning models.

 

 

 

 

Azure Machine Learning cheat-sheet will help to navigate through it.

Conclusion

Data Science is a broad field. It is an exciting field. It is an art. It is a science. In this article, we have just explored the surface of the iceberg. The “hows” will be futile if the “whys” are not known. In the subsequent articles, we will explore the “hows” of machine learning.