Statistical Engineering

Interview with Roger Hoerl, Ron Snee, and Geoff Vining

What is Statistical Engineering (SE)?

How did the idea of Statistical Engineering come about?

Is SE an idea and a practice, or is it a profession, or both? 

How does SE compare to the role/profession of “statistician” and/or “data scientist”?

In today’s episode, Sam Gardner together with Roger, Ron, and Geoff will dive deep into Statistical Engineering and will discuss about the following points:

  • Definition of Statistical Engineering (SE)
  • How did the idea of Statistical Engineering come about
  • Examples of what SE
  • Examples of where it has been applied and if it’s gaining traction
  • Is SE an idea and a practice, or is it a profession, or both?
  • How does SE compare to the role/profession of “statistician” and/or “data scientist”
  • Does SE apply to statisticians working in pharmaceuticals, other Industries, or Government?
  • What is the International Statistical Engineering Association (ISEA)?
  • What are the goals of ISEA and what initiatives are sponsored or supported by the association?
  • Who should join the ISEA (and how to join)?

Listen to this episode and share this with your friends and colleagues!

Roger Hoerl

Donald C. Brate ‘45-Stanley G. Peschel ‘52 Associate Professor of Statistics at Union College

Much of his early research was in the area of regression analysis, especially shrinkage estimators. In the private sector he developed a greater appreciation for, and interest in, experimental design methods. He has recently investigated Big Data analytics, particularly how and why things can go wrong when analyzing massive data sets. This ties to the discipline of statistical engineering, which emphasizes effective integration of multiple statistical and non-statistical methods in an overall approach to scientific inquiry. He is currently conducting research into how statistical engineering can provide effective strategies for attacking Big Data problems.

Publication: https://www.union.edu/sites/default/files/mathematics/202107/publications.pdf

More information: https://www.union.edu/mathematics/faculty-staff/roger-hoerl

Dr. Ronald Snee

Founder and President of Snee Assosciates LLC

He has an outstanding record of leadership in process and organizational improvement in a variety of industries including: pharmaceutical, biotech, clinical diagnostics, chemical, plastics, telecommunications, financial services, newspapers and insurance.  Among Ron’s other achievements, Dr. Snee is credited with leading the design of the first company-wide continuous improvement curriculum for the global giant E.I. DuPont de Nemours. 

He has more than 25 years experience in his field.  Ron holds a host of awards and honors, has co-authored 4 books and published more than 200 articles on process improvement, quality, management and statistics.  He is a recipient of the American Society for Quality’s (ASQ) Shewhart Medal, its highest award, and has also received ASQ’s Grant Medal for his continuous contributions to quality education and research, the American Statistical Association’s Deming Lecture Award and was elected to the International Academy for Quality.

Ron Snee, co-author of three Lean Six Sigma books:

  • Six Sigma Beyond the Factory Floor
  • Leading Six Sigma
  • Statistical Thinking: Improving Business Processes

More information: https://sneeassociates.com/

Geoffrey Vinning

PhD in Statistics, Professor at Virginia Tech

He is a member of American Statistical Association and American Society for Quality and a professor at Virginia Tech. His interests are in the Use of Experimental Designs for Quality Improvement, Response Surface Methodology, Statistical Quality Control, Regression Analysis.

Publications:

  • Vining, G.G. (1998). Statistical Methods for Engineers. Belmont, Ca.: Duxbury Press.
  • Vining, G.G. and Kowalski, S.M. (2006). Statistical Methods for Engineers, 2nd ed., Belmont, Ca.: Brooks/Cole. (2011) 3rd ed., Boston, Ma: Brooks/Cole.
  • Park, S.H. and Vining, G.G., Editors. (1999). Statistical Monitoring and Optimization for Process Control. New York: Marcel Dekker.
  • Montgomery, D.C., Peck, E.A., and Vining, G.G. (2001). Introduction to Linear Regression Analysis, 3rd ed. New York: John Wiley. (2006) 4th ed. (2012) 5th ed. (in press).
  • Myers, R.H., Montgomery, D.C., and Vining, G.G. (2002). Generalized Linear Models with Applications in Engineering and Science, New York: John Wiley.
  • Myers, R.H., Montgomery, D.C., Vining, G.G., and Robinson, T.J. (2011). Generalized Linear Models with Applications in Engineering and Science 2nd ed., New York: John Wiley. 4
  • Does, R.J.M.M., Hoerl, R.W., Kulahci, M., and Vining, G.G. (editors). (2017). Soren Bisgaard’s Contributions to Quality Engineering, Milwaukee, WI: ASQ Quality Press.

More information: https://www.stat.vt.edu/people/stat-faculty/vining-geoff.html

Transcript:

Alexander:   Have you already registered for the live recording of episode 200 on November 30? If not head over to effectivestatistician.com and register there. If you’re already on the email list, then good, you’ll get that invite, Also, are you early in your career? Are you a student? That is more advanced in your career. Then also register for the Workshops that we have. There’s 3 hours of interactive fun workshops that we have to boost your career as a statistician on 1st of December, just head over to the effective statistician and register there. And please tell your colleagues about it as well.

You’re listening to the effective statistician podcast. The weekly podcast with Alexander Schacht, Benjamin Piske and Sam Gardner. Designed to help you reach your potential to lead Great science and serve patients without becoming overwhelmed by work. 

Today, Sam is talking with some amazing people about statistical engineering. So stay tuned for that, and now some music. 

Have you checked out our homepage recently, theeffectivestatistician.com? We are adding their free content on a pretty continuous basis. And we are updating it, and there’s lots of free stuff to find beyond everything about podcast episodes of course, so have a look there and see whether there is something helpful specifically for you. For example, about data visualization, about increasing your influence, or some other nice things that will help you to be a more effective statistician.

I’m producing this podcast in association with PSI. A community dedicated to Leading and promoting the use of Statistics within the healthcare industry for the benefit of patients. Try PSI today to further develop your statistical capabilities with access to the video on demand Content Library.  Free registration to all PSI webinars and much, much more. Head over to the PSI website at PSI web.org to learn more about PSI activities and become a PSI member today.

Sam: Well, today. We are talking with three very special guests. People I’ve known for a while, professionally and they have been involved with promoting the idea of statistical engineering. So we’re going to talk in this Podcast about what statistical engineering is and hear from them, what they think it is. And so let me go through here and introduce who we’ve got with us. So first of all, we have Roger Hoerl, Hi Roger.

Roger:  Hello. 

Sam: Welcome. Thanks for being on the podcast. And we also have Ron Snee. Hello, Ron. 

Ron: Hi

Sam: And we have Geof Vining. So, and while we start here, why don’t you just tell us a little bit about yourself and I guess maybe the two questions that would be good to share would be number one. Tell us how you came to be a statistician? You know that the Reader’s Digest condensed version of that and then tell us what you’re doing right now. So, why don’t we start with Roger, so Roger, tell us about that. 

Roger:  Well, it was not that hard for me. My father was a statistician. He was the first statistician hired by DuPont in 1950. So that was kind of on the back of my mind. As I left high school and went off to college. I was a Math Major. And after I started getting into real math, you know, like real analysis, abstract algebra, that kind of stuff. I quickly realized that I don’t really want to be a pure mathematician. This is not what I want to do with the rest of my life. So knowing that my father was a statistician I thought, well, let me take a statistics course. There was only one statistics course offered at my college, but when I took it, I found that kind of a breath of fresh air that I actually got to see real numbers and solve at least fake problems, If not real problems. And then I did a follow up with one of the professors in regression analysis. And then I really got hooked. So I decided to go to grad school in statistics. So, I spent about 30 years in the private sector. I did two summer internships in grad school at DuPont. Actually, Ron Snee was my boss. That’s how we got to know each other. And in 2012 I decided to make a career change and left the private sector for Academia that Union College in Schenectady, New York, basically, teaching statistics and sponsoring a statistics minor at the college.

Sam: Excellent. Okay, great. Well, thanks for that, and Ron. How about you? How did you get into statistics? And what is it you’re doing now? 

Ron: Well, the beginning is very similar to Roger. I was a Math major at Washington, Jefferson College, as we got to my senior year in math. It wasn’t as interesting and I’d taken the only stat course that was offered. And I kinda liked it. And I decided to go to graduate school and I got an assistantship at Rutgers in statistics. So I took it and so, I was there. Got my, did my graduate work. I was on the faculty for two years then I decided to go to Dupont. And they had an applied statistics group. I was in that group for about 16 years; eight years 

as a technical statistician and another eight years as a manager. Then I moved into some managerial positions in DuPont and later got into Total Quality Management in engineering, and from there I took an early retirement from DuPont and got more involved in process and organizational Improvement working with Joiner Associates, later Bell Atlantic, I was vice president of process Assurance for Bell Atlantic. Then I went to Sigma Breakthrough Technologies. That’s where I got involved in the Six Sigma work so you can see there’s a statistical theme to all of this. And then I moved from there to now Consulting. There’s where I got big time focusing on Pharma and Biotech, and did that for about nine years, then decided to go out on my own and that for the last ten or eleven years. That’s why I’ve been running my own operation focusing on Pharma and Biotech. So that’s what I’m doing today. I teach at Temple University and their Regulatory Affairs Quality Assurance, teaching design of experiments and process monitoring and statistical quality control. I still do Consulting and I do research again on statistical studies in Pharma and biotech.

Sam: Excellent. And then Geoff,  tell us about you a little bit.

Geoff: I’m not a mathematician. So I started out in Chemical Engineering and in fact, my very first, stat course was taught by a former Director at Texacom in the chemical engineering department and I love it. You know, it’s the book. He was the Goodman Wilkes and Hunter book, of course, the first major statistician I ever got to meet with Stu. So it’s kind of sweet. I got interested in some interesting undergraduate. I was there for a long time and became a professional undergrad for a while and  enjoyed life with University Tennessee. So I was very close to a degree in statistics, but did not get my degree and soon thereafter. I got a job as a process engineer making pencils. So, if we’re getting bored today. I could bore everybody even more about how you make those pencils? I’m probably one of the top ten experts still in the world of goods, a lid on because they’re only fine. hahaha. It’s ceramic engineering and it was really a lot of fun. And I was the closest thing that a company had as a statistician to a small company in Middle  Middle Tennessee, who started out in the early 80s in the pit of a dimming audio improvement program. My first wife was from an academic family. So we decided I would pursue a PhD in statistics because I was having fun with it. I came up to VirginiaTech, got my degree and went to Florida for 11 years. You might hear it has always been quality productivity, reliability, generally trying to plan experiments to improve them and I’ve had fun doing it my whole career. So I was acquired for 11 years and we came back here as department head and then once I stepped down as department head it reminds me of my favorite line from Animal House price 7 years of college down the drain so long was at the bar moment ever since then I’ve gotten to play and be able to get in to hasten, some really cool problems. So now, I’m dealing with projects with NASA and Rolls-Royce. Rolls-Royce Motors is high-profile. The NASA process is an interesting one. I’m having fun. 

Sam: Yeah, I think that’s giving that more evidence to that saying that statisticians get to play, and everybody’s play yard. Everybody’s backyard, right? There’s wherever there’s data and wherever there’s problems to be solved. Statisticians can be involved in those areas. So, it’s cool. So today we’re talking about Statistical Engineering, and I think a lot of people would listen to this podcast. Don’t even know what that is. They may have not even heard that term. So I don’t know, one of you probably either Ron or Roger, one of you wants to tell us a little bit about what statistical engineering is.

Roger: Sure. I’ll start off and then Ron can also join in. We do actually have a formal published definition. So if it sounds like I’m reading something, I’m just telling you the formal published definition. Statistical Engineering is the study of the systematic integration of statistical Concepts, methods and tools often with other relevant disciplines to solve important problems sustainably. So what does that actually mean? The terms are chosen carefully, we refer to it as a discipline. That is, it’s the study of something. It’s not a collection of tools, it focuses on integration. So we’re talking about problems that cannot be solved with one method or perhaps even one discipline, but requires integration of multiple methods and discipline in a logical sequence and we’re talking about solving important problems sustainably. So it’s very much problem-oriented as opposed to Tool or technique oriented. So when we talk about statistical engineering, we’re not focusing on a given set of tools. How do we apply these tools? But we’re really talking about how we take a complex problem and think about engineering a solution to that utilizing a variety of tools that might be necessary. 

Sam: Okay, that’s a good thorough formal definition, Ron. Maybe you could tell us little more just if you had to tell someone that you met, you know, casually what statistical engineering as well. How would you describe it to them? 

Ron: Well, I would describe it that we’re used to Statistical Engineering, that when you’re attacking large complex problems and Roger mentioned it, you’ve got to get a solution that is sustainable over time. And so you need a way of doing that and basically Roger hinted at this is statistical engineering is to like Gnostic we use whatever tools are needed and we have no favorite tools or another way of saying it. The only ones that are favorites for us are the ones that work on a particular problem that we are working on. I think the other thing that I would say is that it’s not quite in the definition, although it is it says about the relevant sciences and probably the most relevant science. Well, there’s two relevant Sciences for almost any problem. There is the subject matter of the particular problem, the science of the particular problem and computer science. We think it’s really important to embed the solutions in computer systems, which can help sustain the solution over time. 

Sam: So as you say that and you describe that, I think. Well, that’s a lot of what statisticians do in many ways. A lot of times when they’re involved with solving problems. They do that. And so, how did you come up with this idea of statistical engineering and maybe what distinguishes it from just working in general as a statistician?

Ron: Before we do that. I think it’s important to go back a little bit further. Before Rogers is going to give you an answer to that question. I can see he’s really champing at the bit. 

Sam: Yeah.

Ron: Where this all got started was 2008. Tecna Metrics, Published a panel discussion on the future of industrial statistics and this group identified a lot of problems and there are a lot of problems related to the future but they really didn’t identify any solutions at least as far as Roger and I were concerned. And after we thought about it a while, we felt that the real strategy, the thing we needed to focus on was working on statistical studies, which have a greater impact than what was particularly being done at the time and even still today. And the reason for that is simply, if you want attention, you work on something as important to deliver a good solution. You will get attention. So we went about working on how to go about doing that and came about the idea where we want impactful problems are going 

to be large, complex and non-structured. And so then we invented the whole idea of statistical engineering. The term came as Roger said was a whole idea of engineering a solution and it’s going to be a statistically based solution. So, we published our ideas in a series of papers that came out in quality progress in quality Engineering in 2010. And then it’s kind of going on from there. 

Roger: Yes. I’m going to follow up with some comments from Michael Jordan at Berkley, who gave a talk where he included a slide on statistical engineering and then I’m going to defer to Geoff because he is a Chemical Engineer and we feel the analogy between Chemistry and Chemical Engineering is a good one for the differential between statistics and Statistical Engineering, they’re not competitive, but rather their synergistic, so what I’m going to do, I’m just going to highlight some things that that Michael Jordan said at the Symposium on statistics and the data science era. This is a talk given by the University of Michigan September 20th, 2019. So the title of the slide is statistics as a problem-solving culture. Engineers pride themselves on solving problems. Statisticians don’t think of themselves as being Engineers. We aspire to be scientists discovering truth, but often Society needs us to solve problems to carry out the statistical analog of building a bridge or Electrifying a city. We’re often kidding ourselves regarding discovering the truth. So let’s embrace being engineers and think about what statistical engineering could look like as a counterpart to statistical science. So, rather than studying the fundamental laws of statistical methods or their application, Statistical Engineering is looking at how we can solve these large complex problems, utilizing a variety of methods to statistical and non-statistical to achieve a desired outcome. With that, I’ll pass to Geoff.

Sam:  Yeah, Geoff that would be good to kind of maybe contrast. Like Chemistry versus Chemical Engineering. And maybe that would be a good way to think about it. Then statistics versus statistical engineering.

Geoff:  Amen, that’s where I come in. So it was a passion when we were beginning to set up the association and we were doing research on. The origin of the chemical process industry is actually in the US, the beginnings in Europe where all Specialty Chemicals and their the Chemist work directly with Mechanical Engineers. The Chemist never learned any engineering. Engineers never learned any Chemistry, but they were dealing with very small batch processes. In the U.S. It was often monitored and best provided. So being able to be patient is how you make money. So, on a much larger scale. So actually our origins of chemical engineering has had from MIT from what was called course 10, which was in the chemistry department for many years until he probably spun off as a separate part. And what they were trying to do was to take Engineers, teach them the chemistry that the chemists have enough understanding of the engineering so that they can build a more efficient and effective process. How can we create Building blocks that we didn’t know how to put together? So then chemical engineering people talk about the unit operations, heat transfer, the unit operation, chemical reactor design is a unit operation. And  if you’re doing  distillation or separation processes, it’s all basically the same sum. And the equipment is very simple. So it was this blend of skills for you. So, what it means is, how can we create a strategy for implementing the signs in order to actually make something profitable? And that’s the reason why it’s both sides in engineering. And that’s why I’m very attracted to what we’re trying to do. I’ve been a big believer in what Roger and Ron had been doing since the beginning. So I’ve been happy to be involved in this process. 

Sam: Yeah, I think as I’ve matured in my practice as being a statistician and working solving problems. That idea of the talk that Michael Jordan gave us as statisticians are seeking truth a lot. I think that is a good way to say that even today I was in a meeting and there was another statistician in the room and he was talking a lot about theoretical things. Like what’s the model and what’s new? and what’s Sigma? and that type of thing and that’s not what the people in the room needed. 

Roger: Just to build on what Geoff said. I worked in the Chemical industry at Dupont and then at Hercules, Chemists and Chemical Engineers don’t hate each other. Generally Chemical Engineers respect the deep knowledge in chemistry that Chemists have. Chemists generally respect the processing knowledge of Chemical Engineers who can take fundamental chemistry and figure out how to make money with it. So we want to make it clear. We’re not saying statistical engineering is better than statistics. Statistical science, or is going to replace it. We need statistical science. We just happen to agree with Michael Jordan for a couple hundred years. Our discipline is really focused on going deep deep on science and hasn’t published a whole lot of fraud in terms of the engineering aspect of, how do I take this stuff and actually drive fundamental change with it? 

Geoff: Try to think of having is how to make that ten in. How to get all the parties involved to work together to advance efficient and effective solutions to these complex opportunities that are available around us.

Sam: And Geoff you mentioned briefly there. You said the association and will talk about it a little bit later, but that’s the International Statistical Engineering Association. And when we finish up, we’ll talk about that and what that organization is. So, we talked about it sort of in general. What is statistical engineering? Could one of you give a maybe an example of what a statistical engineering problem or project and solution looks like?

Roger: I think we all can, but I’ll start with one that I worked on. It actually was in finance. I worked at GE Global Research. It was done in conjunction with GE Capital. They had just lost over 125 million dollars on bonds of Worldcom. And we’re trying to figure out, is there a way we could have predicted this ahead of time? And not have to just take these huge losses. So they came to the research center and long story short. We put a team together to look into this problem, whether we could predict defaults in advance. And the first thing we figured out is there is no generally accepted definition of the fault. So different organizations, Dun & Bradstreet, S&P etc. They have different definitions of default. So we realize we’re being asked to predict something that wasn’t defined and then we ask, okay. Well, what data did they have? And they said, well, we’ve kind of been built by acquisition and none of the databases talk to each other. So we don’t really have any data to give you. So we want to predict something it’s not defined with no data and then we get into the issue of how much does a type 1 error versus a type 2 error cost? if we don’t sell and defaults versus if we do so and it don’t doesn’t default. Well, they haven’t really thought through that. There is such a thing as a type 1 and type 2 and how they might be balanced. So we realized we don’t really have a project. We had kind of a mess. We have some Concepts and we had to take a month. Not not one meeting but a fair amount of time just to kind of structure the problem. So that we had something tangible to work on. Now, am I experiencing grad school? And certainly the textbooks that I’ve read. You don’t see this in textbooks and it takes a month to figure out what’s the problem we’re actually trying to solve. And then even when we started getting data, there’s a lot of theory that’s been published on financial default, you know, some people lack and scholes, for example, have one known about prices on it. So we didn’t dive into the data. We dove into the literature to understand the context and understand the nature of this Problem, where it came from, how GE Capital does trading and what their objectives are. So then we spent a lot of time, just trying to understand the background to the problem. Then we had to figure out, okay? Now, how are we going to attack this thing? How are we going to get the right data? What kind of analysis are we going to do? What kind of expertise do we need? We ended up getting a team of statisticians, machine learning types today, they’d be called data scientists, but we didn’t have that term back then, we had pure computer scientists that we’re pulling data from web scraping and things like that and to make a long story short. We ended up with a solution that went through several phases and utilized a lot of different methods from Markov chains to classification and regression trees, the web scraping to smoothing, so there’s quite a few others so it does. It really resembles any of the textbook problems that I’ve read because of the difficulty in getting started in the first place, the breadth of the problem, the personalities involved. We actually had an international team from India, from Stanford, Connecticut and from New York that we had to coordinate. So this is one where we really had to pull together in Lincoln integrate multiple disciplines in order to come up with a viable solution that would actually be sustainable over time. 

Sam: Right? And when you say sustainable, was the output, some sort of prediction system, that would be used on a routine basis. 

Roger: It was, it was, it was an output prediction system that met their requirements, which was, and I’m not joking. It has to be red, yellow and green. It can’t be things like probability default. And so on, it has to be a map with red, yellow and green, because the traders who would actually use it were not people that were necessarily well-versed in statistics. And then we developed a control plan to monitor the system over time and look at type 1 and type  2 error rates to see when this would likely need to be retuned. Realizing sooner or later will need to be retuned. They actually embedded it into their deal process. Meaning the business process for approving large deals, either sale or acquisition, required them to run that entity through our model and include it in a proposal with the model said. So it wasn’t hard wired, but it was included and embedded into the overall business process.

Sam: Right? And so that’s primarily why you needed a broader team because You needed some people, they could actually build that automated system. You know, that some way that was either integrated and another system that when you entered the data, they gave you red, yellow and green. 

Roger: Exactly, right. And building the model was one piece of the overall effort. But a fairly small one. 

Sam: Okay. Well, I think that’s a good example, you know, and I’ve seen that in different areas where I worked, it seems like, that’s a common type of thing that has to happen in areas, like finance and insurance. A lot in particular where you’re doing risk rating. You really have to assess risk and you need to assess it on the Fly. You need like now, I need a measure of risk now. 

Roger: Absolutely. And people are generally reluctant to use Black Box models, where they’re not quite sure why it’s saying they should do what the models are telling them to do. So having some kind of transparency is very important. I think you touched on that this being a somewhat common problem. We do want to make it clear. We’re not saying the rest of the profession is down here. And we are up here, because we’ve thought of all these bigger issues. People have been thinking about these bigger issues for a long time, you know, the George boxes that John Tukey said the world. However, they didn’t really document how they did it. They document how they built the models, but they didn’t document how they approach these problems. So what we’re trying to do with ISEA is fill in those gaps to explain. Then here’s how to think about an approach, these problems that people are facing already, but there aren’t 20 good books on how to do it. Like there are 20 good books on design of experiments, machine learning and so on. 

Sam: Yeah, we were talking before the show a little bit Geoff and I, you know, a lot of the things that go around getting a problem solved don’t have anything to do really with Statistics, but with math, right?

Geoff:  And how do you work with other people in a lot of cases? 

Sam: There’s all those aspects of just general communication, leadership, trust, getting you know, I think what you described in that example, you gave Roger was just defining the problem, well enough. So you could actually get a solution because lots of times, someone just says, can you just fix something? You know, they kind of know what the pain is, but they haven’t defined it in a way. In a structured enough way that you actually can apply a statistical or engineering or computer problem solving method to get an answer. 

Roger: One of my favorite questions. What exactly is the problem We’re trying to solve?

Sam: Yeah, that’s the best question. 

Roger: How many meetings have I been in? That creates an awkward silence in the room because people suddenly realize, they’re not quite sure what problem we’re trying to solve, or they each have their own idea but it is different from what the other people were thinking.

Geoff: It takes three to six months to resolve it because nobody had thought about it. 

Sam: Exactly. Ron, any other examples or things you want to mention about, what does statistical engineering tasks or projects look like? 

Ron: Yeah, and I think the way I’ve been sitting here thinking about one of the audience is the clinical studies and I think it’s a great opportunity for using statistical engineering. And why do I say that? My background, I spent eight years in pharmaceutical development in DuPont and then, actually, I was leading the statistics Organization, doing that work. And then a Consulting job. I also did clinical clinical trials for some of the big Pharma companies. But alright, so statistical engineering. What do you look for?  First thing you look for is that big problem, that important problem and in a pharmaceutical company, You can argue that once you have a pharmaceutical, the biggest problem is getting it developed and getting it to Market. So it’s certainly an important problem. It’s high impact because there’s millions of dollars involved and I think the biggest spend of money is in the form of the development area. So it’s high impact. You look for one where there’s a number of organizations involved and clinical has had a number of different fingers in the pie, both internal to the organization and external and then you bring in a research organization to do these studies and you get another organization. So you got multiple organizations so what happens. Have multiple fingers into pies, multiple agendas and there’s all kinds of politics as a rule of thumb that I use some people ask. Well, when is politics a problem? Well, politics occurs, whenever you have at least two people talking to one another, you will have politics. So, you know, and if you have 20 or 30, it goes out of sight. And then of course you’re looking for something that’s  data-based because we want to use data to solve problems, but But I don’t know, I wouldn’t be afraid to tackle a problem that was important and had a lot of organizations that didn’t have data associated with it, because I’d find some data some place. So those are some of the things that I look for. Roger and Geoff may have some other ideas. 

Sam: Well, and I think in Pharma, you know, like in that area of clinical trials there are lots of problems that are common across clinical trials, right? You always say you are facing the same problem. For example, in the enrollment rate how quickly can you get people enrolled in the clinical trial? And maybe a statistical engineering problem would be, how can we make the process of enrollment more predictable?

Ron: Every clinical group I talked to, that’s one of their problems. How can they do that more effectively? 

Sam: Yeah. And also, how can we make the process of doing that easier as well. What I was talking about today is what we’ve gotta be in a meeting. Seems like now, every time we’ve got a problem to solve every problem solving effort is a new project? And a lot of it has to do because we don’t have good data. We don’t have organized data. We don’t. So it’s always like, we got to go hunt down the data and find where the data is, get it organized. Well, maybe the statistical engineering problem to work on is, how can we have better organized data, and easier retrieval of data? So, it’s ready to  be analyzed, right? 

Geoff: I learned from that experience. And teach others how to do it. That’s what we’re going to do with discipline. with this You know, it’s these one-offs. That’s learning at that stage. 

Sam: So you talked about this. So maybe we could talk a little bit about the ISEA, the International Statistical Engineering Association. Tell me about that organization and its history and what it’s purpose is. 

Roger: Geoff was really the person that spearheaded this effort. So, are  you gonna run now or are we going to defer to you?

Geoff: I appreciate that. I’m a catalyst. That’s how I operate. And I was the point reaction for this. Roger and Ron have been talking about this for a long time and we weren’t seeing Traction in these particular societies, not saying much going on in the American System Association and not much in the American Society for quality. And I’ve been doing some stuff with the military. There was a real opportunity, there’s a conference out there called Data works and which we’ve been engaged in for several years and I began to see that there was a real opportunity to market these two groups that would not traditionally think about as where data analysis was really critical what they did, but they’re getting a very complex problems. For example, my funding for my research here is the science of desk research Consortium Department of Defense, which is all about how we can come up with more efficient weapons tests. Work I’m doing with Rolls-Royce is dealing with the Trent 1000 engine, which they’re losing a billion dollars a year on warranty. So it’s got an impact there and these are complex problems. But you’re getting with people that haven’t been additional statisticians, but who appreciate that there is a need for this. So I saw an opportunity here and I just sent out a feeler. I had an intention that perhaps we could form a society, but I was more interested in the theater, So clearly I reached out to Roger and Ron and several other critical people. Ultimately, we had to group what?  13 or 15. And we met in December, a few years ago. And that was the real origin of the society, but it was the right time at the right moment. We had a critical mass and really good leadership and it’s taken off pretty well. The pandemic is not yielding. I know it’s hurt me because my pay is down to zero, but we still seem to be doing well and Roger, you’re the past Chair. 

Roger: Yeah, I just checked our website. We have 363 members now. Now the organization was actually Incorporated, if I’m not mistaken in 2018 or the end of 2018. 

Geoff: Middle of the  year Roger. 

Roger: So middle of the year 2018 and then something we do that not everybody does, we actually purge our membership roster at the end of the year we send all members an email asking if they want to continue. It’s free by the way, but if they don’t respond, “yes, I’d like to stay as a member”, then they’re purged because people move on. They retire, they pass away. And some organizations have a lot of members, but none of them are really active anymore. So we wanted to make sure our numbers reflect people that as of now want to be members of ISEA and for an organization that’s only been in existence a couple of years. I think 363 is a pretty good number. 

Sam: Yeah. Okay. It does sound pretty good what types of resources or events or things like that does ISEA provide?

Roger: The number one product that we offer to our members, on our members only site, is the ISEA handbook of statistical engineering. So we have a handbook. It’s almost complete. It’s not quite complete, but we have several chapters currently available on the website and we are going to officially roll out the handbook in September. So I’m going to go out on a limb here and say it will be completed in September. The first edition. In addition to that, we have several case studies that are available on the website. As you said, people like to see how this applies to a real problem? So, we have case studies from a diversity of application areas, some in pharma, some in consumer products, and so on, some engineering. And then we also hold meetings. There’s a summit that we hold every year, the summit for 2021 will be online and will be virtual. But Michael Jordan, who I mentioned previously, just accepted our offer to speak at The Summit in November. So more information will be coming out about that. We also offer regular webinars, every couple of months or so. We offer webinars. We just had one actually Stephen Steiner, the University of Waterloo, give a webinar and then there’ll be a next one in September where we’re going to officially roll out the handbook and talk about how people can utilize it. 

Sam: Excellent. So now would be a great time if you wanted to join. The ISEA would be a good time to do it wouldn’t it? 

Roger: You can’t beat the price. 

Sam: That’s right, and I think the website for that is ISEA-change.

Geoff:  ISEA change. 

Roger: ISEA change. (spelling out) I S E A – C H A N G E. O R G. ISEA-change.org  with a dash between ISEA and change. 

Sam: Well, that’s great. I think I encourage any of the listeners here. You have a deeper interest in this you go check it out. And like you said, the cost of Entry is very low so it would be worth it. I’m a member and I’m looking forward to seeing the handbook and the completion and seeing what’s in it. One of the things that we were talking about and I noticed this on the main page and it mentions this a little bit for ISEA that, Is statistical engineering a practice or a profession? Or is it a discipline? You said it’s a discipline but how would you say what it is? That area? 

Roger: I would answer that by saying it is a discipline like chemical engineering as a discipline. Chemical Engineering is not a subset of chemistry. It’s a unique discipline. There is a practice of Statistical Engineering. There’s a lot of practice, a lot of practicing Chemical Engineers, but there’s also a lot of academic Chemical Engineers who do research on chemical engineering. So a discipline has to have a healthy practice and it also has to have a healthy research on a more research base. 

Geoff: So an important component of the discipline is how to collaborate with other disciplines. And I think that’s pretty unique to what we’re trying to do. It’s inherently transdisciplinary and that’s one of the things that we really have to teach in. 

Sam: Has there been any talk of starting like an academic program in statistical engineering at any University? 

Geoff: I just had a course, an intro to discipline engineering, with Diego Coonan? Who’s in Switzerland, is creating a program and has a pretty active program within a business school. We’ve been trying to get that up and we’re going, but the pandemic has had a big impact on that. It’s going to be harder to do it in the US schools. I was headlined in support up here in Virginia Tech. Clock Suey is now here in Industrial Engineering. We’ve been trying to do something collaboratively between statistics, industrial engineering, actually mechanical and Aerospace is really interesting and It’s gonna be backed up by Rolls-Royce projects. I’m getting more traction from the mechanical engineers believe it or not and really from Industrial Engineers because they see the need but we’re nason, we’re learning. 

Sam: It seems to be very cross-disciplinary, and then that’s an even though it is a discipline in and of itself. It really crosses the boundaries of potentially, a lot of other academic and professional disciplines.

Geoff: For now, they communicate across those boundaries. 

Sam: So if someone who’s working as a statistician now, let’s say statisticians in  the pharmaceutical industry and they wanted to start Learning really how to do statistical engineering and start doing it. What would be some good next steps for them? 

Roger: I’d recommend they pull down the first couple of chapters of the handbook from the ISEA website and also go through some of the case studies. There is a pharmaceutical case study there. So they’ll get a sense of how this actually applied to a real problem, especially in Pharma and I’m going to go out on a limb and predict what they’re going to notice is The technical problem of here’s the data. We have to fit a model to it. It will be there, but it will be a fairly small part of the overall effort. Whereas, traditionally with statistical case studies. There’s a one paragraph description of the problem and then 10 pages of data analysis and a model. And very little said after that, about what was actually done with the model, which you developed. So what you’ll notice in these,  there is the technical problem, the data, the modeling, that’s part of it, but you’ll see a lot more discussion of how do we figure out what is the problem we’re trying to solve? Where did the data come from? What were the limitations? Once we had a model? What do we do with it? How do we convince people that it should be used and figure out where to use it? Work with those organizations to get it used and so on. So I think they’ll see we’re talking about something broader than perhaps  what they’re used to with a statistical case study. 

Geoff: This is a case to case study. The whole idea was to showcase the theory,here ideally, we’re showcasing the sustainable solution that came out of this whole process. 

Sam: Excellent. Okay, well I agree and I get it. It seems like you keep developing a lot of good resources there for people to use and that would be a good place to start as they go to the ISEA. 

Ron: Think there’s one other thing that needs to be added. 

Sam: What’s that Ron? 

Roger: We’re on the edge of our chairs Ron. 

Ron: The thing that needs to be added is that it’s good to do what Roger said, but I wouldn’t do that very long without going in my organization and looking for an important problem. That needs a solution. And what you’re going to find is, once you find it. It’s going to be a mess as Roger said, so it’s gonna need to be structured. You, as a statistician, can see all kinds of things that you don’t know how to attack. Get some answers off the website and so forth, you can talk with people, but I think the best way to learn how to do it. Is to do it. 

Sam: And just struggle with it too. right? Struggle in doing it. 

Geoff: And we record the journey. Yeah, we got to sell. We have to talk about the journey because that’s where the learning is. 

Ron: Yeah. Oh, yeah. 

Sam: I think that’s the great knowledge sharing within an organization and even across organizations about that aspect of the practice of what we do. Maybe that’s some time lacking, and maybe it goes back to what we said before that Statisticians are seekers of truth. Right? So what they want is I’ve got this really awesome statistical method that I implemented that I developed.

Roger: And but really the best is, it’s better than anything else. Regardless of what problem we’re trying to solve. 

Sam: Right, right. 

Geoff: And this here is telling the Journey about how we got these solutions so that other people can build off of that. Yeah, saying I see it more. As much as anything else is being able to tell the story about spotting your case studies pretty well. 

Sam: That’s excellent. Well, I think you guys have given a really, really great overview of what statistical engineering is. I’m really hopeful going forward. This becomes more noticed as disciplined, right? More people, maybe even people you start seeing people having a job title as Statistical Engineer. Yeah, that would be a big step forward if this is going to move forward as a really a new discipline that people follow.

Geoff: As a group, that is statistical engineered. And it’s been in existence for about 10 years. 

Sam: Oh? See? Either you guys are already there and some place and my I have not come across that in my experience, but, you know, mostly in Pharma, but I wonder too if there’s a little bit of, just concerned about using the word engineer, for someone who doesn’t have a formal degree in engineering as well. You know, so that’s another thing to think about. 

Roger: Well, I think if you make a job title out of it, that becomes an issue. For engineering as we know it is accredited by Abet. Organizations like that. However, I have seen people with job titles Financial Engineer, Data Engineer, and to the best of my knowledge. Those are not Abet accredited programs.

Geoff: All the engineers that’s already in. Yeah. 

Sam: Quality engineers. And you can get a certification in quality engineering through other organizations, right? I guess that, that is something that wouldn’t, I, guess you shouldn’t be. We shouldn’t be afraid of taking that on as a title, if we wanted to. 

Geoff: Coming up with the Department of statistical engineering would lead to some of those very questions. But, you know, creating the discipline and job titles. I don’t think it is as big of an issue and look at NASA. I mean, they have a group in statistical engineering, of course, that’s a very Engineering dominated agency of the government. 

Sam: Exactly. 

Geoff: And they’re comfortable with that. 

Sam: Well, that’s great. Well, I think what we’ll do is we’ll wrap this up. I mean, we probably could talk a lot more about just examples and what the future holds for statistical engineering, but I really appreciate you all taking the time to come on and talk about what it is and tell us about the, the ISEA and really appreciative too of the initiative that you’ve taken to move this forward. And it may be that this is the solution for the ongoing debate about our statisticians data scientists and our data scientist statisticians. Maybe if we just started talking about statistical engineering, we wouldn’t be fussing about those distinctions as much. 

Roger: Yeah. I’ve kind of come to a new place whenever Masters questions and I basically say, “look, I don’t care what you call it”. If you’re doing this stuff, you’re going to succeed. So you can call it whatever you want. But just make sure you’re doing these things, like very carefully documenting the problem you’re trying to solve, studying the background and the context to the problem, developing an overall strategy for how you’re going to attack it, ensuring that solutions are sustainable, and they’re actually put into place. If you’re doing that stuff. Let’s not debate what to call it. We wanted to come up with a formal definition because a lot of these other terms are not well defined today. A lot of Journals use the terms AI, machine learning, data science interchangeably and the public doesn’t really know what any of them are. 

Geoff: I think anybody knows what they are. You wanna know the truth. 

Sam: Yeah. Alright. Well, we’ll wrap this up and this is a great encouragement for those who are listening to move forward, do good science, but also do well at the process of solving problems in the organizations that were involved in. Yeah, thanks to all  of you.

Geoff: Thanks Sam for doing this. This was great! 

Ron: Thanks for having us.

Alexander: This show was created in association with PSI. Thanks to Reine who helps us with the show in the background and thank you for listening. Head over to theeffectivestatistician.com to find the show notes and there’s much more material to improve your career and have more fun at work. Reach your potential, Lead Great Science and serve patients. Just Be an Effective statistician. 

Never miss an episode of The Effective Statistician

Join hundreds of your peers and subscribe to get our latest updates by email!

Get the shownotes of our podcast episodes plus tips and tricks to increase your impact at work to boost your career!

We won't send you spam. Unsubscribe at any time. Powered by ConvertKit