 Deuxième demi-journée de notre rencontre horizon. Je suis très heureux d'accueillir à Robert Rodriguez et j'aimerais que, si vous appréciez, Bob, vous introduisiez en français parce que je ne veux pas oublier de remercier Kevin Bleakley pour donner une excellente traduction pour vous dans l'anglais. Donc, Robert Rodriguez est Seigneur Directeur du Statistical Research & Development de SASS Institute qui est le leader mondial du logiciel statistique. Il est titulaire d'un PhD de Statistique de l'Université de Caroline du Nord à Chappell Hill. Avant de rejoindre SASS en 1983, il a été chercheur au General Motors Regers Laboratory et puis il est fellow de l'American Statistical Association dont il a été le président en 2012 et il est très important de le noter, past president pendant l'année 2013, qui était l'année internationale de la Statistique. Et je ne crois pas le trahir en disant qu'il a été l'un des promoteurs les plus ardents de notre discipline pendant cette année. Et donc, c'est très naturellement un très grand plaisir pour nous et pour moi en particulier de le recevoir aujourd'hui. So Bob, it's time for you to come, so I leave you the floor. Thank you Jean-Michel. My bio sounds much better in French than it does in English, even though I didn't understand it. I have been looking forward to this day ever since Jean-Michel invited me and it's really an honor to be here because the French Society of Statistics has done so much to promote our profession during the past year. And on behalf of your sister society, the American Statistical Association, I want to congratulate you on everything you have done to organize this special event. At SAS headquarters where I work in Kerry, North Carolina, we have celebrated during the past year, the international year, with these banners up and down the streets so that our visitors can see that statistics is a very important part of our corporate heritage. And now that the year is over, the banners have been taken down but we're left with some very important questions which I want to talk about this afternoon with you. This is a picture of a road, of a journey ahead and I want to talk about where our profession is headed. Now I know that there are students who will be watching this on YouTube and many of you teach in universities. So there is a parallel question that I want to consider with you today and that is where can a statistical career take you? And there really has never been a better set of answers for our students. And I want to focus my presentation on the sectors of business and industry complementing what you have heard earlier this morning and what you will hear later today. So my focus is going to be primarily on opportunities that I have seen in the worlds of business industry and also to some extent government and research. I want to use this metaphor of a journey for our profession and every journey that we take really has three parts to it. The first is the anticipation of a journey. We think about what lies ahead and today I want to give you five trends that are shaping our future things that we should be thinking about as we look ahead. Then there is the preparation. What is it that we pack? What do we put into our suitcase or our travel bag so that we will be well equipped for the future? And I want to spend time thinking with you about training for success in some emerging areas of statistical practice and in particular I want to talk about the importance of statistical leadership for our profession because I truly believe that leadership is our most critical resource and perhaps our most underdeveloped resource. And then third, there is the destination. Where will we arrive? And I strongly believe that if we anticipate correctly and if we prepare correctly that we will arrive at a destination that provides greater visibility for what we do as a profession and also provides us with new roles for statisticians in these new areas of practice. So I will start with the anticipation for the journey and I want to give you five trends or new developments that have been shaping the future of the statistics profession. The first of these trends is actually probably the oldest of the five and that is the growth of this area that we call business analytics. Now what I have learned in talking to statistical audiences over the past few years is that there are many statisticians who are not very familiar with this area so I would like to begin with a brief definition. I think of it as the application of multiple areas including statistics but also drawing from data mining, forecasting, optimization and increasingly text analytics to make critical business decisions and to add value to what the organization does. Now my favorite example of adding value through statistics to a business comes from a company in Scandinavia that operates the bridge that connects Denmark and Sweden and perhaps some of you have been to that part of the world and have traveled over this bridge, the Oresund. Now this company cannot really raise to the tolls on the bridge what they charge to go across the bridge so in fact the only really viable way in which they can pay off the construction loan for this project is really to increase the frequency with which people travel across the bridge. And in order to accomplish that these companies have been developing statistical models using surveys and demographic data and these models predict what the preferred activities are of their customers for various segments of customers and that allows them to offer and provide these customers with special offers which give them discounts for activities on the other side of the bridge. So for example if you live in Denmark and you like to eat gourmet food then you are offered discounts on restaurants in Sweden and these discounts and these offers have turned out to be very popular and in fact they are increasing the frequency with which people travel across the bridge. Well business analytics is in fact used in many many different sectors including government and other areas. A few other examples are a reduction of adverse effects associated with the drugs. There's a lot of data that's collected on that in the pharmaceutical industry. In banks people are using business analytics to study the patterns of customers so that they can do a better job of retaining customers over their lifetime. In the hospitality industry building better relationships with customers so that they will come back to their favorite hotels. And increasingly the most critical problems are characterized by massive amounts of data typically data on customers or patients and transactions as well as new sources of data and the need for rapid decision making. Now business analytics is important to us because it is really driving the demand for statistical skills and this is happening globally. More and more business organizations have been using statistical and related skills and techniques to compete on analytics and there are many good books published on this topic. What we're finding is that in survey after survey company executives say that they are very interested in increasing their analytical talent. This is something that we've been very interested in exploring in the American Statistical Association and we have been reaching out to the business world to understand how we can better prepare our students to meet their needs. The problem I see is that we are doubly disconnected from this world of business analytics. On the one hand the world of business analytics prizes value statistics as a tool as a set of skills but they don't really recognize statistics as a profession. They see many other people as capable of using statistical tools and they're hiring them and so statistics as a field has relatively little visibility in this arena. At the same time I think our profession has often overlooked business analytics as an opportunity a major opportunity for statistical practice even though many of our students are now moving into these areas and finding very successful careers. So one of the things we need to do is to understand the goals, the language of business and we also need to better prepare our students to solve the new kinds of problems that they will encounter in these areas. Now a related trend and this is a newer one is something we've been hearing about now for several years and this is this phenomenon of big data. I like to think of big data with three words that begin with R. The three R's of big data. The first is relentless. Big data is generated on a continuous basis. Much of it comes from online transactions among people or transactions among people in online systems or in other cases it's coming from sensor-enabled equipment such as sensors on jet engines or expensive turbines. Secondly big data is relatable and what really gives it its importance and its value in practice is that it can be related. It can be linked to provide very detailed information typically about customers or about people and their patterns. And that is what allows big data or large amounts of data in the business world to be used to create these new business opportunities and many of you of course are familiar with software such as Google for instance, search software that is very powerful, very useful and has really led to the creation of new business models. And then third big data is renowned it's famous. I don't think a day goes by but I don't see an article somewhere or someone sends me a clipping about some new aspect of big data. And certainly in the media many parts of of our society United States government for instance it is perceived increasingly as the answer to the big questions so we're reading, we're hearing a great deal about big data. Now the good news for us is the big data is creating a shortage of statistical problem solvers and McKinsey Global Institute tells us that in the United States alone by 2018 we will face a shortage of somewhere between 140,000 and 190,000 people with skills in statistical methods and data analysis. Now the problem is that we are simply not training enough students to meet that huge need. In the United States the total number of degrees that we confer annually in statistics is somewhere around 4,000. The number of PhDs has been fairly flat across the years. The number of master's degrees is increasing and the number of bachelor's degrees is increasing and that group is increasing because students are arriving now from high schools to American universities understanding that statistics is an important skill and that it can be used as the basis for a very useful and in demand type of career. This is a very, very important trend. Now there is still this huge gap we are not going to be able to meet it and we need to really be thinking about how that gap is going to be filled and already it is starting to be filled at least in the United States by people who loosely turn themselves data scientists. Data scientists are not trained currently in graduate programs in American universities although there is a lot of thinking going on in the funding agencies for the US government and in many universities about whether we should create programs in data science and so there is a move in that direction. But we need to think about what really describes all these people who are now billing themselves, describing themselves as data scientists and are finding that they are very much in demand in many companies. If you talk to these people or you talk to their employers what you find is that data scientists are regarded as innovative problem solvers, people who can learn from data. They have expertise in building statistical models and using machine learning. They have generally a very good understanding of the business problems of the problem domain. They are supposed to be good communicators of what they learn and they are also very well known for having specialized or advanced programming skills they call themselves hackers. This is a description that if you factor out machine learning and the specialized programming skills what you are left with is a description of a well trained statistician and in fact that has been the goal for many many years to produce people with these types of skills. So now we face competition from people who are outside of our circles and the question is what are we going to do about it? How do we respond to that? Well, related questions are how many statisticians are really trained in all of these areas and how many more could we train and another question is is it really true that data scientists are equally strong in all of these different areas? Well I have been talking to people and reading up on data science. There was an interesting study published last year they did a survey of people who call themselves data scientists and they found that in fact they're not all the same they fall into different categories just like statisticians fall into different categories in our profession. Some data scientists focus more on understanding the value of a data project to the organization. Some focus more on cleaning up the data getting it ready for analysis or storing and managing the data focus much more on understanding complex processes. So there are different categories of data science what we find from this study is that statistics is actually important to all of the people in these categories and what we also find out is that data scientists typically are very deep in one of these areas and they have ability in all of the other areas, some ability. So we'll keep that in mind and I'll come back to that point when we talk about skills for statisticians and how we should be shaping our students. But I think there's an important lesson here to be learned from these people who are emerging as data scientists. We have the challenge of explaining to others including the media university administrators people in national funding agencies what is it that sets statisticians apart what makes us different from other people who can learn from data. I think primarily in terms of three ways in which we should really be distinguished of course we find features in data we are good at that, we are trained in that but in addition we are trained to guard against false discovery bias and confounding and false discovery was something that you heard a little bit about earlier today from Emmanuel that that's an important issue in the science area well it is also important in other areas and statisticians are trained in how to deal with that. Like other people who are learning from data we are capable of building models that can explain, that can predict, that can forecast but in addition we also qualify their use through measures of uncertainty and through thinking very hard about the assumptions that go into these models and then third like others that work with these enormous amounts of data in business and industry we are trained to work with observational studies but in addition statisticians have always added a great deal of value by producing designing studies that produce data with the right information because without the right information content no amount of big data is going to suffice to answer the big questions. There is a fourth point that I would like to make that sets us apart and that is that we as statisticians contribute as a profession and our journey really has to be as a profession. We cannot undertake this as individuals and what being a professional statistician means is that you don't just learn a set of tools but that you participate in a community that has a legacy of theory, research, continuing education, accreditation ethics and above all community and so these are the things that bind us together and give us strength for the journey. Trend number four is the need for distributed computing and again you heard about that this morning the fact that more and more data is being stored and managed across multiple computers and that introduces special computational problems for statisticians and this is something that we need to be anticipating. The increased size of data means that it cannot be stored in traditional databases on a single computer and so we're seeing more and more use of distributed computing platforms and massively parallel databases and the fundamental problem and Emmanuel explained that very well is that interaction of the data, moving it from these distributed databases into computational environments is really a bottleneck problem and so the challenge more and more is to co-locate the analysis with the data or to find very, very innovative techniques that avoid having to move the data. So the expectations for distributed computing using an analysis using that type of distributed data those are growing and so we really need to prepare for distributed computing. Well what are the advantages of this approach? Well we gain performance, the ability to tackle larger and larger problems by distributing the data across the nodes in a cluster of computers and we also gain performance by breaking the work into tasks that can be done in parallel by the computational power, the memory and the computational resources that are available on those nodes. And what's important to note for us is that even though we may not be using these systems today the price of the hardware keeps dropping and dropping and so there's a very good chance that we will see more of this type of data more and more of these resources available in our computing centers and our data centers and that people will come to us from all kinds of areas with problems and with data that are stored and available in this fashion. So that is very much a trend that we have to anticipate. Well that means in turn that we will need to think about statistics on a big scale. We have to think about scaling our computations understanding where the data are located and what proportion of the work can be done in parallel. We have to understand how to scale our programming techniques. We need to be able to apply parallel programming where it offers benefits and that means in turn that our students and some of us will need to pick up some new programming skills. Above all, I think we have to be able to scale our thinking. Do we understand the mechanisms that generate the data. Statisticians always excel at asking those kinds of questions and understanding that the importance of that question. Understanding whether the data are suitable for answering the big questions and understanding what constitutes a useful model for that type of data and how do we visualize the models and how do we visualize the results. So those are all challenges for us as we move into this era of big data. Trend number five is the prevalence of unstructured data. You're looking at a graphic that was used to advertise a White House workshop on big data that was held back in May of last year. And if you look closely at this you can see that there are many terms here that really relate to unstructured data and the use of text analytics. The reason for that is that when we think of big data we should really be thinking not only of the size of the data but also the variety of the data so much data really is in the form of email or documents or perhaps images and the experts in information technology say that roughly 70% of enterprise data is in this form. So if we're going to flourish in this new world we have to anticipate being able to work with unstructured data as well as the structured data that we're currently trained to analyze. We need to learn more about text analytics which is really nothing more than statistical learning applied to collections of documents and that involves understanding the relationships between documents and terms and combining free form text with quantitative variables and there are techniques that ultimately convert things over or reduce them into a form that can then be handled using the types of techniques that we're trained to use. And some examples of this type of work are again uncovering adverse effects of drugs by looking at notes taken by patients and doctors or identifying fraudulent insurance claims or understanding where there are early warnings of customer dissatisfaction from call center data. All of these are ways in which textual data is being used and where there's increasing opportunity for statisticians to get involved with that type of data. Well I've given you five trends that I have seen ranging from this new area of business analytics and the arrival of big data the emergence of data scientists the need for contributing computing and the prevalence of textual data. These are all things that we should be thinking about as we look ahead at the journey for our profession. The next question I want to take up is how do we prepare for the journey what are the skills that we will need to have and how do we equip our students for success in these areas in the future. Well I think we need to really think about how we shape our students for success et what we mean by shape is best explained using these little analogies with letters of the alphabet. Traditionally statisticians have been trained to be very deep in theory and methodology and we have really been immersed in what I think of as the core statistical skills and that's what we emphasize currently in our programs and we're very good at that and so our students in a sense are shaped by eye I mean they're very very deep just like the letter eye is long and narrow. What we've seen from understanding the popularity of data sciences is that they are not eye shaped they are T shaped what that means is that they have they have good statistical skills but they're also very competent in computing and also in communication I believe that more and more that's what we ought to think of as the model for students and the way we train them in the field of statistics. We need to continue giving them the strong core skills but we also need to prepare them with computing skills and increasingly with communication skills because that's really what is in demand out there and that is what will help them to succeed. In addition we'll need to be what I call pie shaped. Pie shaped means deep in statistics and also deep in computing and somewhat competent in other areas and so I think there's a need for an opportunity for people who can go that extra step and become pie shaped. Now for those of you who teach in university programs a very related question is what do you give up and there's only a finite number of hours with the students and there's much to teach them. Those are hard questions and in the United States we're beginning to think very hard about that and many of our departments are really beginning to innovate and revise their curricula and it's a very important contribution and I think it really does need to start with the universities based again on an understanding of what it is that their students will succeed in the future. So I think the answer is T shaped with some emphasis in some departments I think on preparing students who are pie shaped and I think they will really be in demand. I'd like to make five recommendations for those of you who teach in universities and these are not based on my personal beliefs but based on many conversations that I've had with people in parts of the world both in the academic world and in the sectors of business, government and industry. So these are recommendations that really grow out of perceptions that have been brought to me in my interactions with many different groups. I think we need to equip students to solve problems with new types of data in many cases types of data that we haven't seen or maybe that are not yet there a problem that involves new data. We need to update the curricula to meet the needs of industry and government. There are many, many career opportunities for students but we have to understand what it is that will help them most to succeed. We need to increase the number of undergraduate degree programs and that is starting to happen in the United States. Traditionally the field of statistics has been thought of as a graduate field. When I started as a student you didn't major in statistics. You went to graduate school in statistics and most of us in my era began with an undergraduate degree in mathematics. We will still need to have mathematical skills as part of our core training but I think there's our field will really not flourish until we have a strong undergraduate component just as you find in so many other areas of science. I think about chemistry, physics and so many other areas where they succeed because they equip students at the undergraduate level whether or not they go on to a graduate degree. I think we need to train more secondary level statistics teachers. In the United States we have a program in high school called advanced placement statistics which enables students to take a course that's equivalent to a one semester university course and they can gain credit for that and this is a program that has really grown in popularity. It is the fastest growing of all of the advanced placement subjects that are offered in American high schools and last year 167,000 students took that exam. You compare that number with the number of people who graduate with university degrees and statistics to see that at the high school level or the secondary school level we have an important opportunity that we should not be neglecting and that is to invest in good statistics teaching so that future generations of students in every field will leave school with an appreciation for the power of statistical thinking. And then last I think we need to emphasize career success skills at the university level and what I mean by that is these skills like computing communication learning how to give a good presentation and again we do have programs in the United States that are starting to do that that are offering special courses and I will say a little bit about that later. Within the American statistical association we have developed a number of resources for all of our members really some of these are especially for people in universities who are beginning to think about how to revise their curriculum. A task force that worked during 2012 put together a series of recommendations for master's degree programs that were developed with input from executives in business and in government and also input from former students recent graduates who were asked the question what is it that you wish you had learned in your master's program that most would have helped you and if you're interested in the answers those are available on the ASA website. We have introduced training for ASA members in how to give effective presentations that communicate the relevance and the impact of our work. We introduced in 2012 a new conference series called the statistical practice conference which is very different from other conferences that we have in the sense that it really emphasizes practical skills and experience as opposed to focusing on research. We have very strong conferences for people who are interested in research presenting their research work or learning about new methodology but we did not have anything that would really serve the needs of people in areas of practice where they could benefit from hearing about the experience of others. We have some initiatives that are currently underway that are being led by the current president of the ASA. One has to do with big data. We are reaching out again to stakeholders and these are executives in companies to learn more about their perception of data science and what they would like to see statisticians coming to work with in terms of skills and abilities. We are working on a new program to help to train high school statistics teachers and we are also developing training in leadership skills for statisticians. I want to spend a few minutes talking about leadership because I feel very strongly that leadership is our most underdeveloped resource. It's so important for us to think about statisticians as leaders. Fundamentally and I think this conference or this event really demonstrates this fact very powerfully statistics is truly an interdisciplinary endeavor and the growth the future success of the field of statistics really depends on getting people in other areas to understand what we contribute and then to act upon the results of our contributions. What this means is that we need the ability to communicate to others the importance, the impact of what we do and increasingly play a leadership role in all kinds of organizations and settings where we are going to be put in contact with people who are coming in with new problems and new types of data and there is a tremendous opportunity for statisticians to act as leaders. In principle statistical leadership in my mind is all about influence. The ability to influence people in the organization help others to succeed through our contributions. It also means increasing the visibility of our profession by influencing the perceptions of what we do. So that is in a theoretical sense what I think statistical leadership is all about but in practice I think what it means is introducing statistical knowledge early in the process where we have been traditionally which is to be brought in often late or after the fact. Most statisticians who work in research organizations have been faced with this problem of not being involved early enough in a research study in terms of the design and often we are left with having to fix up the data after it has been collected. Very, very traditional type of problem and so we have an opportunity we always act as leaders in terms of introducing our knowledge early in the process and then drawing others into it. Very hard to do in many organizations but I can tell you that in the business world in the pharmaceutical industry in particular the people at the top are really saying we want statisticians to be playing the role of a leader in all these new areas where we have new data coming in. So there is a huge opportunity for us. Well then what kinds of leaders do we need to have? Now I ask that question to a number of my peers people who have served as president or former presidents of the American Statistical Association and of course I could have asked many other people but what I really wanted to do is just get some input from people that I have worked with and the reason I ask the question is that all too often in our field when somebody speaks of a leader in the field of statistics we think of the people who are the intellectual giants and all of us know who they are these are the people whose work we put up on a pedestal we admire greatly but we all too often don't think of ourselves as potential leaders we think of leaders as being the people who are creating the new advances in the field or perhaps the people that we elect as presidents of our societies but we don't often think of ourselves as being leaders so here are some I ask these former presidents what do you think of leadership where do we need leadership what kinds of leaders do we need to have and here are their answers we need leaders in all areas in small activities and in large activities at every level of work that we do and I think that's a very very important answer in 2009 statisticians really need to be the voice of statistics in their organizations we need leaders with skills Stu Hunter who is an extraordinary communicator reminds us that we need to people who can go out and communicate the enjoyment of statistics and show people that statistics is fun that it's interesting and that it's important that it makes a difference Dick Schaefer reminds us that we also need these organizational and planning skills and certainly without those skills we would not have an event like horizons of statistics today and we need leaders with a very broad perspective John Kettenring reminds us that we need leaders who understand and perceive the opportunities that lie at the interfaces of statistics and other fields and again I think this event the importance of that perception and because those are really what ultimately those connections that strengthen our profession and its impact on society now I could say a lot more about how you learn to become a leader some departments in the United States are actually offering courses in leadership, they're optional courses I tend to think of leadership as something that can be certainly learned through training but most of us don't have the opportunity to do that in graduate school so I think really leadership ability for many of us has to be caught rather than taught and what I mean by that is that we learn about good leadership by working with good leaders and following their example and we're better than to learn good leadership than by serving in a professional association like the French statistical society or the American statistical association or any other organizations that we have throughout the world and this is where you will find if you're a young person good role models and great examples of people who are leaders and who demonstrate their leadership by serving others so if you're a younger person or if you're just wondering how do you become a statistical leader I think you really just begin by volunteering to serve a need whether it's in your organization or in your society but everyone who's a great leader has told me the same thing over and over and over again that they simply began not by thinking of themselves as a leader but by seeing a need and going after ways in which to fill that need I think you grow as a leader by working on your communication skills by practicing your writing skills by becoming a better writer a better speaker and by continually you succeed as a leader by serving others so there is my speech on leadership I think it is a tremendously important aspect of what we could be doing there's so many other things we need to learn but having more of us think about becoming leaders I think will make a huge difference for our profession in the future well I have covered the anticipation of the journey and I've talked about the preparation for the journey in some ways in which we're thinking about that in the United States and perhaps you're thinking along similar lines here in France I want to finish by giving you some thoughts about the destination for our journey and where we can end up if in fact we learn to anticipate and prepare ourselves or our students for these major new needs that are coming up over the horizon I want to begin with a comparison of ways in which we can contribute in these new areas of emerging statistical practice and contrast them with the important ways in which we continue to contribute in what I think is more the traditional areas of statistical application and practice in a practical sense what many applied statisticians have to do today is access data from relational databases and many in our profession work with that type of data in the future we will increasingly be confronted with distributed data and distributed computing and so we'll have to start to switch over that but that's the new need that's the new way in which we can contribute we have always had to prepare conventional data for analysis I don't think anybody with real experience in statistical work has been able to get away from the problem of taking the raw data and getting it into the form that it can really be analyzed effectively with statistical tools so we've always had that problem of preparing conventional data I think the data scientists are discovering that problem for themselves now it's a big one increasingly I think in statistics we will have to learn how to the preparation will involve integrating data from different kinds of sources and some of it may be textual data, some of it may be more structured data we have always excelled at designing studies or surveys or experiments that structure the data in such a way that the data deliver a maximal information content for the analysis that is when to continue I think to be one of our strengths but I think we're going to be increasingly having to think about drawing from I've got these I've got these cells switched here we will have to be able to draw from work more and more with observational data and use a variety of tools here like statistics, data mining forecasting, optimization and so forth we're going to have to be able to think more and more about the problem up front of the as opposed to being the people who analyze the data at the end of the day so I think that will be a switch we'll I think continue to be experts in these core statistical areas statistical inference that's what we do but I think the challenge will also be to work with people in other areas and combine ideas from data mining, from forecasting and optimization and so forth so I think these are new ways in which we'll find we can contribute in these new areas of practice and I think we'll start to see more of a transition towards those areas but I think we'll continue to rely on these strengths that you see on the left hand side of the page new career opportunities and this is important if you have students or if you're a student or a young person who is watching today's event currently in traditional areas of practice for example in pharmaceutical companies there's very much a culture in which statisticians are viewed as professionals high level managers understand that we're part of a profession and they support the participation of their employees in statistical professional activities in the emerging areas of practice that I see that is not the case and that's a concern for me the dilemma there is that executives managers view people with statistical skills as very important, very valuable but they value people for the skills and they don't think about the profession in these areas we'll have to build a professional presence and we'll have to do that by reaching out as professional societies and connecting more and more with those groups and that's one of the things we're doing or we're trying to do within the American statistical association traditionally we have worked and this is a classic model for how a statistician works as a consultant providing consulting support for other people who own the data and who are the domain experts but again I think the and I'm hearing this especially from people in the pharmaceutical industry from the food and drug administration who are saying we really need statisticians to act as leaders to come in very early on and kind of take charge to define the problem formulated statistically and then involve others in the solution traditionally we have acted in the role of supporting decision making for products for services or processes in these new areas I think we'll be expected to we have the opportunity to show the way in terms of how new sources of data can be exploited to advance the high level goals of the organization and then lastly in terms of communication I think we have another challenge and another opportunity there with good communication skills will be in a position not simply to give presentations to our peers but also to give presentations that show the value of what we do and to customers to and to managers so those are some comparisons of course these are all approximate but I think you can see the opportunities and the new roles for statisticians I want to summarize now and give you what I think are the most important points I've made we're in 2014 and I think we see this paradox which is that in these new areas business analytics big data there's this enormous demand for statistical tools and people who can apply the tools great career opportunities but at the same time the concern for us is low visibility for statistics as a field and that could be higher and so we need to kind of work on that the future of our field I think depends on how well we serve the needs in these new emerging areas of practice and how well we equip our students to solve problems with new data sources and we have to really develop the ability to communicate the relevance the impact of our contributions and also to lead and influence within our organizations I'd like to leave you with some personal applications I hope you've heard some things that cause you to think about how you might train your students differently or how you might innovate the curriculum in your university or how you might approach your work differently or how you might communicate differently with say the media or with people in the business and industry world but here are the three applications I would most likely to most likely to leave you with first of all it's a challenging journey and that means that we need to prepare second it's a professional journey it's not a journey that we can take by ourselves and what that means is that all of us should participate in the profession and if you're a young person or a student what better place to do that than in an organization like the French Statistical Society and the third point I want to make about the journey is that it will be a rewarding journey and so you should persist thank you very much thank you both for this night talk there is some time for questions please feel free to ask it in French or in English directly if it in French you will provide some translation Kevin some questions thank you I'm Stephen Besso from University sorry hello thank you I'm Stephen Besso from University Paris 13 Nutrinette which time frame do you see for the journey you describe I think you heard the question what do I think in terms of this the time frame I think it's not too soon to think in terms of starting now how long will it take before we get to the point where we see the difference probably I can't be specific in terms of a number of years but I think we're already starting to see in the United States University administrations talking about creating departments of data science and I think we can't wait until those are in place and then start thinking about how we react I think all too often our profession has waited and then we have wondered why people in other fields they seem to now own an area that we think of is something that we should be leading some of the things that I mentioned that can be done in a year or two the initiatives that I described that we're doing in the ASA are actually one year initiatives the master's degree recommendations those were put in place within one year it took us several years to create the new practice conference that I mentioned but now that's an annual event and that is starting to grow we have now one year initiative in place to create leadership training and we'll be offering that this summer for the first time at the joint statistical meetings and we'll certainly start small that's always the case but the leadership training is pretty much being developed within one year I mentioned these as examples just to show that some things don't require five years whereas other things will certainly take longer to answer the question over here two times maybe more in your presentation you refer to big questions but what is a big question and how do you know whether a question is big or not and who what persons, what organizations are entitled to ask big questions this is a play on words but of course we've got so many people talking about big data I really enjoy your question though it's a great question a couple of years ago I was invited to during 2012 I was invited to Washington by the Obama administration they launched a new announcement of a new big data investment program and this really involved a lot of the sciences especially physical and biological sciences and there was a lot of talk about the big scientific questions and how all the sciences are inundated with data and are looking to to answer the big questions in their fields with these large amounts of data and that for me is where I sort of picked up this phrase the big questions don't give you many different answers but when I read the media what the media says about big data they're often talking about solving very important scientific questions questions that other speakers are in a much better position to address than I am but I use the phrase big question to challenge statisticians to really think about the information content in the data and whether it's sufficient to answer the questions which I use the phrase but to answer your question the big questions are the ones that are being posed by other people with large research budgets another big or small question you were the leader of the American statistical association you were the leader so my question is very simple how many statisticians are there in the US and how many belong to a statistical association oh that's a very that's a excellent question I can only answer half of it we have roughly 19,000 members in the American statistical association I know for a fact that there are many people in who are teaching statistics in universities who are actually not ASA members and I wish we could reach out and connect with them and include them we are we are also what really concerns me though is the number of students who are student members of the ASA who are students only for a short period of time and then once they take a job especially if they take a job in one of these new areas they tend to drop out of the ASA and then we don't recapture them I don't have any way of answering your question precisely about how many practicing statisticians are out there I would have to believe that it's multiples of the membership of ASA but we really we should be spending more time thinking about especially how do we retain the students we would like to keep as young professional statisticians and build the ranks and I suspect if we did that successfully the size of the ASA would probably I think we would see tremendous growth if we found ways to do that it's time to thank thank you very much Bob for this talk