Simon J. Sheather is Professor and Academic Director of Texas A&M's Master of Science in Analytics, which offers courses from the top-ranked Mays Business School in Houston, Texas. He brings a wide scope of experience in management and the integration of analytics into organizations and businesses.
From 2005-2014, Dr. Sheather served as Professor and Statistics Department Head. Previously, he was a faculty member at the Australian Graduate School of Management at the University of New South Wales.
icrunchdata speaks with academic leaders about their programs and latest initiatives. Today, Dr. Sheather touches on the beauty of modern predictive modeling, the big data "illusion of accuracy" and the future of Analytics education.
Let's dive in...
That’s a very interesting question; it was in the middle of 2004. I went on a small vacation and when I returned to the office, sitting on my desk was a letter from Texas A&M inviting me to be a candidate for the role of Department Head in Department of Statistics. It was completely out of the blue; so naturally I thought about this and ultimately ended up applying.
I interviewed in October with other candidates and ended up being the sole finalist and was notified in November, 2004. I reached a deal with the then Dean of College of Science, Dr. Joseph Newton, to become the Head of Statistics at A&M University. I arrived to take up my position in March, 2005.
We welcomed our first cohort in the fall of 2013. We started with some thrill seekers, because we received official permission from the Texas Higher Education Coordinating Board six days prior to classes starting. In order to get the program approved, there were a whole series of levels that we had to go through. We had 13 in the first cohort.
Our MS Analytics program is a partnership between the Department of Statistics and the Mays Business School. It’s taught out of the Mays Business School facility at Houston CityCentre, and it's beamed live online, with half of the students in Houston and the other half across the country.
As Department Head, I started the MS Statistics online program in 2007. We would record the classes and put them up on the web and students watched the recorded lectures and submitted the homework and took the tests. The marketplace kept asking for data science, statistics and business – the ability to produce predictive models to solve business problems.
SAS approached us with the idea of doing a distance based big data/analytics program. We thought about it and what we decided was not to do something completely out of distance; we needed a live audience in order to have interaction with the class. We visited a number of schools, in particular, the program at North Carolina State. Dr. Michael Rappa, the Director of that program, was very generous with his time and they shared their curriculum with us.
The big picture is that students need to be working in a job that’s related to analytics and data. We require at least three years of work experience. We’re not very fussy about what degree or background the student has. We only require one statistics class.
What’s very important is that they have support and mentorship from their company because a key part of the program is a five semester capstone project where students solve a business problem from their own organization. The support and mentorship that each student must have is a crucial factor.
While students are in the program, they get exposure to a lot of case studies. We talk with lots of companies and offer them this deal: If they will share with us a dataset and a business problem, we’ll use it as a case study in class, and we’ll share the students’ solutions at the end of semester. That’s quite popular because it gives another perspective, it gives a different set of eyes looking at the problem; but the key thing about our program is that our students get to apply what they learn every week to their major capstone project. When they learn a new technique, they can look at their own data and apply that technique to that data.
There’s been a bunch; but a lot of them have been solved. The first problem was: Can we capture the data from customers in real time? Yes. Then, can we aggregate and combine data from different sources? And the latest one is: Can we build models?
In the old days we used models for customer segments. People with similar demographics and similar behaviors were all modeled as if they were identical. The beauty of modern predictive models is that we can build a model for each individual customer. We can customize the model to each person. We don’t predict what happens to a group – but what happens to each individual, and that is very exciting.
You really need both. You can be as smooth as silk with the soft skills, have no content – and have a problem. You can have amazing content and not being able to explain it to others and still have a problem.
We’re very proud that our program has a great balance between the two sets of skills. The student learns the ability to solve problems using data and the ability to communicate the solutions.
One size doesn’t fit all. Look around, there are many programs in the country offering a different focus to certain needs and backgrounds. We’re focused on solving business problems and applying statistical modeling to solving those problems.
We want to train folks who will run analytics teams in the future, who will sit with senior management teams involving decision making on new products or offerings. We’re training those people. If you want to be that person, this program is of great fit.
What gets me excited is solving real business problems. We use an example in class – which was the class project in predictive modeling where we looked at Airbnb. What drives the ratings in apartments in New York, San Francisco and in Paris? There were about 10 factors, and the most important one was customer ratings on bathroom cleanliness. I found that to be particular surprising and going back to the topic about where analytics is heading is that it reinforces management's views of what’s important; and, it usually adds one or two other secret sauce ingredients that they haven’t been aware of.
Analytics education, especially our program, teaches students pursuing degrees like this one, to find that secret sauce and make better business decisions.
That’s a perfect description of what my mother wrote about me. What I’m interested in right now is what’s called one of the illusions with big data. For some of the standard statistical methods, you need to be cautious when using, especially if the data is really, really big.
For instance, I have a paper coming out that looks at New York City taxi fare data and another one that looks at airfares with Southwest Airlines. I look at hundreds of thousands of taxi fares and airfares, and what I’m interested in is, what affects price? You come up with a model and get estimates and then measurements of precision. Is it $2.50 plus or minus how many cents? And some of the standard estimates of precision are way too small giving what people have started to call this “illusion of accuracy” that doesn’t quite exist. Now what you do to sort this out is something I’m thinking about right now.
I wanted to be a high school mathematics teacher, so I started a mathematics degree. One of the advisors told me there was this new subject called statistics. You should take that if you want to be a math teacher, so I did. I ended up taking two classes from a very famous Australian statistician called Peter Hall, who sadly died earlier this year. He was an incredible educator and really changed my whole life.
I ended up getting a degree in both mathematics and statistics. I got offered a job with the Australian Bureau of Statistics, and I could’ve gone to Canberra, but instead I went to graduate school. I switched universities and went to La Trobe University in Melbourne, Australia and got a PhD there.
Don’t miss any Essendon winning games, mate. Don’t be hard on yourself, and work more on life/work balance.
There’s a few fun facts about me. In 1975, I came in first for the state of Victoria in the Latin competition, while in high school. When I was a college student, I taught people to play tennis in order to help me pay for college. That was my first teaching job. It was frustrating at times, but it made teaching statistics later in life much more straight forward.
Hard work is a great competitive advantage no matter how smart or how intelligent you are. Hard work is an incredible competitive advantage.
Excellent words to live by. Dr. Sheather, thanks for taking the time and sharing your perspective.