ASA Connect

 View Only
Expand all | Collapse all

What you didn't learn in school

  • 1.  What you didn't learn in school

    Posted 12-04-2015 11:54

    Hello Everyone,

    This forum has a lot of interesting discussion with good input from many people, so I wanted to get your thoughts on the following question, please. What didn't you learn in school that you wished you had?

    Our profession is changing in "The Real World" and quite rapidly today. We always have to be learning new things on the job, but are there foundational skills that you would have been better off learning in college (high school!?) but you didn't? I imagine the question would be answered differently depending on when you graduated and what you do now. Since colleges and universities are trying to adapt their curriculum to what they perceive to be the needs of employers, more recent graduates might have a different answer than those of us in the "Silver Circle."

    From my own experiences, leaving college 20 or 30 years ago depending on how you look at it, I have my own opinions, but I'm very interested in what the group says. I also see a lot of resumes and job descriptions, so my opinions may be slanted more to those experiences. However, here are the top few things I wish I had learned in school that I could have used at Kellogg's and, then, with the many clients we work with in our Business Intelligence and Analytics Practice. (And maybe they tried to teach me these things and I was just too bull-headed to learn. :-)

    Integrating with the client -- this is like collaboration, but more so. It also could be described as learning to live in their world. Statistics was always an us and them, even when we talked about teams and such. There could have been more about getting into their heads, their objectives, their language.

    Get it right vs. get it done -- clients have, more often than not, cared more about getting something that's in the right direction for what they need (60 - 80% of the way there) than for something that's 100% statistically accurate. Some of that requires my educating them, but I also have to compromise with them in ways that are going to give results that are quick, actionable and still minimize risk.

    Business presentations -- How to tell the story of the data in such a way that the audience (business, engineers, product developers, VP's, etc.) will not be lost in the technical details, feel confident in the results, and know how to take action in their jobs. Everything I saw and did (or as I remember it) in school was very technical and that transition was hard.

    Project management and software development -- Not to be a project manager or software developer, but to have some of those skills. My only project management in school was getting the homework done by Friday. It would have been good to learn more about requirements gathering, task estimation, status reporting, change control, error checking, etc.

    That list is the top things on my mind now to get the discussion going. What do you all think?

    ------------------------------
    Chuck Kincaid
    Engagement Director
    Experis Business Intelligence and Analytics
    chuck.kincaid@experis.com
    ------------------------------


  • 2.  RE: What you didn't learn in school

    Posted 12-07-2015 07:57

    I think learning something about Database systems is important. Too many professional statisticians use their stats software to perform database functions. That's like driving a Smart Car while towing yacht, then complaining about how slow it goes.

    If statisticians knew how bad the algorithms are for sorting and joining data in R, SAS, Stata, etc, versus SQL Server or Oracle or Hadoop, they would feel foolish. Since they don't know about Database software, they don't use it. 

    ------------------------------
    Andrew Ekstrom



  • 3.  RE: What you didn't learn in school

    Posted 12-09-2015 09:34

    I agree that statisticians could benefit from learning more about database systems. I'd add that learning more about databases, independent of database systems, would be a good first step. Data and database independence mean that researchers and statisticians can share data and databases without having to license, learn, or rely on a specific database system.

    As an example of what one might learn from studying databases, consider so-called "normalization" of a large data file with hundreds of variables and repeating groups of variables. Decomposing this large file into several or more tables linked by distinct key columns in each table minimize structural missing values, expose duplicates, generally reduce the data footprint in storage, sometimes dramatically, speed up searches and links to other databases, and make it possible to restructure data to fit many different analytic procedures (one row per entity, multiple events per entity, hierarchical, etc.). Or take these simple but useful ideas from database theory: normalization moves information from variable names into data, scanning of many narrow rows in a database works much quicker than reading of equivalent information from many variables in a row of a table, indexes have many advantages over sorting, ordering of columns, and other physical rearrangements of files, and arrays are dangerous in that they have preset maximum dimensions while data streams don't.

    As a demonstration of an indexed JOIN of two very narrow databases using one of those supposedly clumsy statistical software packages (SAS):

    options compress=y;
    DATA rLarge ;  /* Stream a Billion Obs. */
         DO key = 1 TO 1e9;
             value = ROUND(1e9 * RANUNI(3131313),1.);
            OUTPUT;
         END;
    RUN;
    %put >>>>>>>>> ;
    PROC SQL _method; /* Find Extremes Fast. */
      CREATE TABLE extremes AS
      SELECT * FROM WORK.rLarge
      /* Criteria */ WHERE value BETWEEN 0 AND 100
                 OR value BETWEEN (1e9 - 100) and 1e9  
      ;
    QUIT;

    DATA rSmall;  /* Stream a Million Obs. */
         DO key = 1 TO 1e6;
             value = ROUND(1e6 * RANUNI(717171717),1.);
            OUTPUT;
         END;
    RUN;
    PROC SQL _method ;
      create table rJoin as
      select r1.key, r1.value
      from WORK.rLarge as r1 INNER JOIN WORK.rSmall as r2
           on r1.value = r2.value
      ;
    QUIT;

    For those who don't have the SAS System installed, I'll summarize the results. Creating table rLarge takes between ~3 and 10 minutes on my modest Windows desktop with an SSD. I should add that the PC is multitasking while running the program. Creating table rSmall takes considerably less than a minute. The JOIN optimizes to a hash index search and takes about the same time as it takes to scan rLarge.

    I've had to learn many lessons from theoretical statisticians while building research databases. It seems only fair that statisticians would have to learn some of the basics of database theory when analyzing observations in large databases;>

    S

    ------------------------------
    Sigurd Hermansen



  • 4.  RE: What you didn't learn in school

    Posted 12-07-2015 07:58

    I'll comment more about what I've seen missing from new-hires, as I came out of grad school the first time 35 years ago.  Statistical fundamentals:  I think too much emphasis is placed on applied techniques.  New-hires are more inclined to make assumptions that allow them to use the analysis approach they used in school as opposed to examining the phenomena they are working with and adjusting methodology appropriately.  Strategic Planning:  This is not so much one to get in initial education, but it becomes critical as you move up the ladder.  Once in the working world, few statisticians seek training on this subject, so they tend to muddle through somewhat haphazardly.

    ------------------------------
    Patrick Flanagan



  • 5.  RE: What you didn't learn in school

    Posted 12-08-2015 10:20

    ​​To my mind the most important thing to learn is HOW to learn!

    Robert

    Robert C. Elston,                                                            
    Distinguished University Professor Emeritus    Email:  robert.elston@cwru.edu                         
    Case Western Reserve University                    Website: http://darwin.cwru.edu/rce/  
            
    3000 Galloway Ridge, Apt G301                      Phone: (001) 919-542-0841
    Pittsboro NC 27312

       






  • 6.  RE: What you didn't learn in school

    Posted 12-07-2015 09:20

    That is a fine list.

    I would add Nobody cares about what you know unless they know that you care. As a consumer insights professional I did not control any  product or campaign under study - just the study itself. That meant that I did not "own" the ultimate success or failure in the market. When I learned to reframe my interest as an employee and colleague with shared goals overall, it was easier to collaborate with my clients  to design rigorous tests and even deliver disappointing news.

    ------------------------------
    Regis OConnor
    President
    O'Connor Analytics LLC



  • 7.  RE: What you didn't learn in school

    Posted 12-07-2015 10:26

    As someone who is out of grad school for little over 5 years I'm willing to bet that I am not in the 'Silver Circle' yet (not sure what the year requirement is actually) but I'm happy to offer my thoughts :-).

    It may come as a surprise, but I am in agreement with all the posters here.

    1. Textbook examples. A big lesson for me was that those fabulous textbook examples rarely work right 'out of the box' and that you rarely find real world data that fits the textbook method. Sometimes I think those textbook data sets were made specifically for the analytical method.

    2. Statistical negotiation skills. As Charles said, I learned pretty quickly that clients rarely want to implement a full statistical design. I get the question all the time "Can you help me design a robust study...but we are not interested in statistical significance". To add to that, almost rarely (in my experience) can the sponsor afford the fully powered sample size so there is always this back and forth on how to retain the statistical validity of the trial but minimize sample size. It would have been nice to learn how to play this back and forth between statistician and client.

    3. Team playing/integration.  This is closely related to the negotiation skills. In grad school my view of stats was: someone slides some data under our door, we write the analysis, we slide said analysis back under the door. I never realize how integrated a statistician needs to be (i.e. from clinical trial development phase to execution). It is not a lone wolf profession.

    4. Software programming/coding. I never expected to be coding as heavily as I do. Back to the textbook examples, coming out of grad school, I expected all the data that I will be analyzing will be given in a nice tidy little package. However I find that 90% of the 'statistical analysis' is spent getting the data in the correct format. After that its easy peasy (relatively speaking).

    5. Report writing/presentation. Given that we are in such a technical field we are taught how to write and present in a technical manner. This works great in grad school because the people you are presenting to is fellow statisticians and/or stats professors! I got so used to assuming people knew what a p-value or a linear regression beta coefficient was, that I took that for granted when I first arrived at work. I quickly learned the power of the KISS principle.

    Cheers,

    ------------------------------
    George Skountrianos
    Statistician
    Hollister Incorporated



  • 8.  RE: What you didn't learn in school

    Posted 12-07-2015 10:26

    Hi Chuck - 

     
    What a great question!  
     
    I really like the business side / soft skill side items you mentioned already.  
     
    I also want to add - data cleaning.  As a student and as a professor, I received / distributed clean datasets for students to analyze, but that's just not "real world."  I think for professors, readying data to be slightly messy, but not too messy that the student drowns, is a balance and time-consuming.
     
    Thanks for compiling this!
    Melissa Kovacs
     
     
    ------------------------------
    Melissa Kovacs
    Principal / Founder
    FirstEval, LLC



  • 9.  RE: What you didn't learn in school

    Posted 12-07-2015 12:20

    You present a great question, Chuck! This is actually something I have been pondering quite a bit lately as I approach 3 full years working/ being out of school.

    You made some great points as to what you wish you had learned. I especially liked your comments about your desire to have been taught more communication skills (integrating with the client, business presentations). My background is in research in social science (Experimental Psychology) and I am thankful that the program I attended stressed communicating well with one's audience in order to make the data mean anything useful (dissemination is a large focus in experimental psych). I found this to actually be more important when I transitioned from academia to the business world because I work with folks across many backgrounds of experience and focus, not just other researchers. For me, continued development of strong communication skills has been just as important as my knowledge about research and statistics. This has also helped me in respect to your comment about "Get it right vs. get it done" - I have had many opportunities to discuss project needs internally and with clients and determine what can be done now vs. what can be done in the future (or with different data). In these situations, it's important that we make sure we can surpass technical jargon (from both sides) and all get on the same page. I hope that schools and teachers continue to development instruction around communicating statistics and research across audiences.

    I really wish I had learned more about database systems and software development (coding) in school. While I was instructed on SPSS syntax and that was extremely valuable, I lacked programming skills coming out of school that would have greatly enhanced my ability to work with data more efficiently. The first skill I was tasked with learning at my first full-time job after completing my degree was SQL. Upon learning SQL, I couldn't believe I had never been exposed to database systems and languages before, because it was so helpful and efficient! I use those skills almost daily in my current position, even doing a lot of data selection and cleanup within the database system prior to importing data into my analytical software for analysis. (I do also use SPSS syntax! That is still very useful). Learning software development takes more time and effort. That is still part of my development plan and I wish I had more time to learn the skills right away! I am considering taking an intro to programming course to advance my learning at a quicker pace.

    ------------------------------
    M. Baylee Moore
    Statistical Research Analyst
    Imprezzio, Inc.



  • 10.  RE: What you didn't learn in school

    Posted 12-08-2015 11:08

    Chuck, this is a lovely question indeed.

    The reality is that school will never fully prepare us for what we need to do in the real world. One reason for this is that the expectations we face in our working environments continue to evolve and increase in complexity as we progress in our careers. Another reason is that some teachers may miss the opportunity to teach not just their subject, but also valuable life skills by either modelling those skills in front of the students or mentoring the students outside of class.

    But what schools can do is to nurture our ability to learn, prepare us to some degree for the transition into the real world and help build our confidence as professionals. 

    A few years back, I used to teach a statistics course to graduate students. They were all extremely bright and talented but many of them seemed unaware of their individual strengths and were low on confidence. The competitiveness of the school environment had the unintended consequence of making many of the students feel "less than" some of their peers. I worked with the students in a way that reinforced that everyone is appreciated for having unique talents, strengths and skills. I also encouraged all students to leverage the talents, strengths and skills of their peers when choosing project teams.

    It would be great if schools could find the right balance between competitiveness and collaboration. Competitiveness is about the individual whereas collaboration is about the greater good. 

    It would also help if schools had a formal program in place for mentoring graduates as they seek their first position. Personally, I felt "abandoned" after graduation - and was even told that I was someone who was no longer the responsibility of my school, as everyone was too busy with current students. This was a harsh experience and I can only hope that things have improved since then. The success of a school is intimately connected to the success of its students, so the school should be invested in making sure its graduates go on to bigger and better things. Making the fresh graduates aware of mentoring resources they can tap into either at the department or university level is paramount.

    In terms of hiring, I think many of us can attest to the importance of having been given a chance somewhere along the way by someone who believed we could achieve more if given the opportunity.

    Sadly, as the economy evolved, the concept of extensive training on the job became rather obsolete. Back in the 60's, a new hire could expect to get a variety of training courses in the first few months and years on the job. Nowadays, a new hire has to hit the ground running on Day 1.  This places tremendous pressure on new graduates, as they are expected to posses a laundry list of skills which require years to acquire. 

    So, if I were to look at a fresh graduate's resume, I would seek signs that they have the ability to learn things on the fly and are not afraid to tackle challenging tasks. I would also look at their ability to collaborate with other people and whether or not they have shown early signs of leadership.  

    The key is to look at each potential hire not as as a "final product" but someone who could become a valuable asset to the company/organization with the proper amount of training and mentoring.

    ------------------------------
    Isabella Ghement
    Ghement Statistical Consulting Company Ltd.



  • 11.  RE: What you didn't learn in school

    Posted 12-09-2015 09:55

    One comment that I haven't seen is the ability to ask questions to get to the underlying need.  I often find that clients/collaborators will come with a specific question that does not address the underlying need.  We end up playing "20 questions" to circle in on the question that they are trying to address and also to get to a question that can, in fact, be answered with the data at hand.  This is particularly the case with secondary data analysis.

    ------------------------------
    Lawrence Muhlbaier
    Associate Professor
    Duke University



  • 12.  RE: What you didn't learn in school

    Posted 12-09-2015 09:35

    Others have already hit on this somewhat, but the three things my colleague Sam Huddleston and I addressed in a recent article in Analytics Magazine (and OR/MS Today) are listed below.  Coincidentally, the article was titled, "Not taught in school (but useful in the 'real world)."

    1. The need to fully embrace the second half of the famous quote from George Box.
    2. How to communicate the results of your analysis to decision-makers.
    3. The importance of creating a good visualization of your data/model (a subset of the previous task).

    You can read the full article at: http://www.analytics-magazine.org/september-october-2015/1392-0viewpoint-not-taught-in-school-but-useful-in-the-real-world

    ------------------------------
    Scott Nestler



  • 13.  RE: What you didn't learn in school

    Posted 12-09-2015 09:55

    I would say that understanding that we don't know it all and getting a job means continual learning.  Difficulties that are found in datasets from particular field are unlikely to be unique.  Cast a wide net in your literature search.

    clients provide messy datasets.  New statisticians need to provide detailed directions on formats.  Also developing a set of auditing algorithms really helps because datasets always contain errors,  I can often spend more time preparing the data for analysis than completing the analysis itself. 

    Find a way to have fun and relax yiour mind.   Find and learn from more experienced colleagues.  Be patience with job promotions.  Your time will come.

    -

    ------------------------------
    Steven Correia
    Environmental analyst (Retired)



  • 14.  RE: What you didn't learn in school

    Posted 12-09-2015 09:56

    I'd encourage you to look at the "Report of the ASA Workgroup on Master’s Degrees" that was published, I believe, as an Amstat News Article in Feb 2013. I have shared these contents, which are online at http://magazine.amstat.org/wp-content/uploads/2013an/masterworkgroup.pdf, for the past few years in a seminar with our M.S. candidates. Opening the seminar by asking students what they think companies are looking for in new hires, and what companies think are missing in new hires, is a good exercise. The report, which surveyed such companies, includes "the answers"--something that students always love to have!

    ------------------------------
    Joseph Voelkel
    Rochester Institute of Technology



  • 15.  RE: What you didn't learn in school

    Posted 12-09-2015 10:28

    Hey, Chuck!

    Nice topic and many nice points.

    I would add a couple more...

    Metadata: Real world data almost always has an undocumented or partially documented history. That history is often extremely important for analysis and understanding. I find it very challenging to argue for more resources applied to that documentation, more effort to communicate that history as part of data sharing,  and more detail added to the now readily available appendices of published research.

    And metadata has an umbrella topic...

    Non-statistical uncertainty: Samplers have long discussed, but it extends to every act of inference and indeed summary description. Computational power is giving us more tools to explore it, but thinking about and communicating non-statistical uncertainty is still a growth industry.

    ------------------------------
    Mike Hoekstra
    Mathematical statistician
    Centers for Disease Control and Prevention



  • 16.  RE: What you didn't learn in school

    Posted 12-09-2015 10:55

    I love that someone pointed out that students need to learn how to learn. I'm always suspicious of a job candidate with straight A's as many of those I have interviewed achieved that academic recognition by taking unchallenging classes. The B student who struggled has often turned out to be the better performer compared to the one who sailed through by taking the path of least resistance.

    Students need to learn to excel even when doing things they do not enjoy doing. The absolute coolest job I ever had was about 25-30% stuff I didn't like doing. The worst job I ever had was 80-90% stuff I didn't like doing. I still was expected to do those things well. I think that students can practice this skill in those courses they are required to take but would rather avoid.

    At least one person mentioned that there is too much emphasis on techniques and not enough on fundamentals. This is a problem in every discipline. Too many engineers learn how to use a specific drafting or FEA package but nothing about drafting or theory of finite elements. They don't know a wrong answer when it punches them in the nose. I design airplane parts, sometimes knowing that the "answer" you have is wrong can be more important than knowing the "correct" answer.

    ------------------------------
    Douglas Hall
    Sr. Principal Engineer
    Honeywell