Aspiring Statisticians: Advice for Career & Education

[3.141592]

Hello,

I am afraid this thread is a bit cheeky: I am not a statistician and have never had job working with data, but I am a mature student, studying a part-time undergraduate in mathematics and statistics and I would like to work as a statistician.

If any members would like to share their experiences of anything related to this subject I would be tremendously grateful.

I mean anything from course content or academic skills you found most useful in the workplace; the working conditions; what jobs one can do whilst studying to improve chances of getting a statistician's job upon graduation; if one is already taking the full catalogue of statistics courses, what mathematics courses are most relevant etc.

[I have done lots of research from various sources on the web so this is not a lazy attempt to get information without looking, but I thought it would be a good addition to the carefully constructed stuff on career, company and professional-body websites.]

For my part, I have zero interest in finance, and the idea of marketing or business is quite dull to me, but less so than finance. What does appeal is the idea of medical statistics (trials, epidemiology), being involved in the discovery of new knowledge for worthwhile causes. Operational research sounds very varied. Actuary is out (for me) as I am in the UK and have bad A-levels. But please include anything relevant as hopefully this will be a thread for all budding statisticians.

Many thanks in advance to anyone who contributes.

 

[MarneMath]

Statisticians, by the very nature of the job vary from degree of education and work done. In businesses, I've seen people who just happened to remember their stats course fall into the role of the local stat person. Through some self-study and funded education by the company end up being fairly competent individual who can shift through some data and be a useful member of the team.

Course Content: Keep in mind, I am not a UK student, so I don't know if there exist a stats degree or if it falls under a math degree with an emphasis on stats. Nevertheless, I do suspect you'll take 1 course that studies probability using calculus and distrubitions, with a second course on basic statistical interference. You should be able to take a course on experiment design and maybe some time series and varies other stat focus courses. If you want to focus on medical side, I highly recommend focusing on experiment design.

At the graduate level, things become more interesting. During your undergraduate years, it would behoove you to take a course that introduces you to measure theory. It will become the language of good deal of your courses, at least it was for me. It's easy to get lost in the woods of theory at graduate level stats, so I encourage you to stay focus and ask yourself constantly how a topic can be used in "real life."

Work: I started out doing financial research. A lot of measurements, a lot of data reading, a lot of second guessing people. I ended up working in epidemiology. In both fields, knowing SAS and R and a couple of other high level programming languages is an assumed skill. I run and make models constantly. Testing, modifying, and banging my head against wall because my code monkeys have failed me! Because I'm mostly tilted towards bioinformatics, I work with a lot of Variable order Martingales

Expect a lot of people coming up to you and asking you to help them with their stream of data and interpert the information. Sometimes you can do it, but a lot of time, people misunderstand stats and think "hey I got a bunch of data here, make it mean something!" Most of the time, I just have to say, I'm sorry the way the experiment was conducted or the information I have here isn't enough for me say anything meaningful. Thus, I now take a more proactive approach. I hold bi-weekly meetings with people who are conducting experiments so me and my little team can provide input through out the whole entire process instead of being consulted at the end.

All in all, I enjoy sitting in my corner cube and doing my job. I'm never the "leading" expert on the field, and I'll never become the star researcher or guy whose name people will know outside the office. Nevertheless, within my little cubicle world, I'm known as the reliable guy who will give you a straight answer. I like it like that. I can't 'fix' a lot of problems, but I generally involved with everything, even when it isn't my main focus. This translates into a lot of meetings. I had a conference call on new years day =(. Still, I like the scientist I work for, and it's nice to be useful. Plus, my wife's friends assume I'm a genius working on some top secret government project or doing important work, even if all I do at times is shot notes into a trash can :).

 

[StatGuy2000]

Hi there. I would concur with everything MarneMath has just posted. I would also suggest taking courses in applied statistics (if they are offered in your school), and if you intend to pursue graduate studies in statistics, to enroll in a course on statistical consulting (this was a specialized course offered at my alma mater to MS and PhD students to provide first-hand experience on various projects that are conducted at the Statistical Consulting Service -- most statistics departments in Canada, the US and the UK will have some equivalent course available). The consulting course is especially useful in providing you with first-hand experience interacting with clients, who typically come from many different areas (clients who I've interacted include members of the Zoology department and researchers from a teaching hospital affiliated with my university).

As far as my own work, I currently work as a biostatistician focusing on the design and analysis of clinical trials. As MarneMath as suggested, a strong understanding of both SAS and R is an assumed skill. Another important skill is being able to communicate the results of your analysis to other stakeholders who may or may not have an understanding of statistics in a clear and concise manner. In my own role as a biostatistician I frequently interact with physicians, pharmacists, data managers, programmers, and those from regulatory affairs on questions related to the specific trials I'm involved with and knowing how to communicate the results of the analysis is crucial.

I hope my post here is helpful. Please feel free to PM me if you have any further questions.

 

[3.141592]

Thankyou MarneMath for such a detailed reply, it's very helpful.

I don't know if there is anything meaty in my course on experimental design (but I am hoping so!) but I know there is a chunk on medical stats which includes clinical trials so perhaps it comes up there.

I don't know what 'measure theory' is right now, so I better look that up! I am sure you are right about looking constantly for real world uses (well, uses I am interested in, since it's all useful some way or another); whenever I get stuck and frustrated in mathematics it's always ten times worse if I can't see a relevant-to-my-interests use for what I'm battling to understand.

I read that high-level statistics and probability allow one (to some extent) to flit between industries much more than many other jobs might. Since you began doing statistical work in finance and are now in epidemiology, I wonder if that speaks to your experience? Actually, epidemiology interests me a lot but it looks like one needs higher degrees to break into that field(?).

Your comment reminds me of that quote (Fisher) about a statistician conducting a post mortem of an experiment - perhaps being able to say what it died of! I do like the idea of being a 'go to' guy in a team actually, so long as it is something that requires skill, rather than from laziness i.e. being the 'go to' guy for fetching more paper for the printer... (been there).

Is it reasonably common, if you can say, to be involved with experiments or trials or let's just say a project, from conception all the way through to recommendations based on results? I think I would really enjoy that, rather than getting something on my desk with no idea how it got there, tinkering a bit, then shipping it off with no clue what becomes of it or my efforts. Quite alienating I have found. Also, I don't want a job where I am purely a technical guy, spending all day shuffling symbols on paper or screen. Would you say there is much room for lateral thinking or creativity when figuring out how to collect the data a project requires?

Your description of having a rep as a solid player with stats skills, who is involved in projects from start to end, sounds just wonderful.

Thanks a lot.

 

[3.141592]

Hello StatGuy,

Thanks very much for replying. I am based in the UK and studying with the Open University. This means I am a distance learner. I don't see an MSc in my near future, mainly because of money. There is only one university in the UK that provides a statistics MSc via distance learning. I believe this MSc has a module that casts you in the role of a professional statistician. I also believe they have a consulting unit you can shadow and, of course, the master's project is almost always brought in from industry.

Unfortunately that course will elude me for a long time. Might you have some suggestions as to the sort of jobs one could get before having the degree? I have a degree already but it is arts based and non-numerate. I have no qualms with taking a non-graduate, low-paid job, if it is a foot in the door for a data job and will be at all relevant on my CV come graduation.

Can I ask if you had/needed much background by way of biology for your biostatistician work? From job postings I've read that seems like one needs at least a medical stats degree if not outright biological knowledge, public health etc. rather than being from just a maths/stats background.

The design and analysis of clinical trials sounds to me to be wonderful. Involved in the discovery of new knowledge, perhaps 'myth busting' many nonsense claims from self-proclaimed health gurus about why you need to buy their latest yogurt to give you eternal life etc. and all for a socially-worthwhile cause. I wonder if I am glorifying it a bit though.

R and SAS are assumed knowledge? Oh dear. On my degree I will have courses using Minitab, Winbugs, SPSS, GenStat, Mathcad and Maxima. Are these not very widely used? I thought SPSS was - or not in medical areas? I don't mind learning in my own time for a job, but since all my free time is spent on the degree (I am a part-time learner) I don't really have room to spare right now. Perhaps over the summers.

Thanks again and thankyou for the offer of a PM.

 

[StatGuy2000]

Hello StatGuy,

Thanks very much for replying. I am based in the UK and studying with the Open University. This means I am a distance learner. I don't see an MSc in my near future, mainly because of money. There is only one university in the UK that provides a statistics MSc via distance learning. I believe this MSc has a module that casts you in the role of a professional statistician. I also believe they have a consulting unit you can shadow and, of course, the master's project is almost always brought in from industry.

Unfortunately that course will elude me for a long time. Might you have some suggestions as to the sort of jobs one could get before having the degree? I have a degree already but it is arts based and non-numerate. I have no qualms with taking a non-graduate, low-paid job, if it is a foot in the door for a data job and will be at all relevant on my CV come graduation.

Can I ask if you had/needed much background by way of biology for your biostatistician work? From job postings I've read that seems like one needs at least a medical stats degree if not outright biological knowledge, public health etc. rather than being from just a maths/stats background.

The design and analysis of clinical trials sounds to me to be wonderful. Involved in the discovery of new knowledge, perhaps 'myth busting' many nonsense claims from self-proclaimed health gurus about why you need to buy their latest yogurt to give you eternal life etc. and all for a socially-worthwhile cause. I wonder if I am glorifying it a bit though.

R and SAS are assumed knowledge? Oh dear. On my degree I will have courses using Minitab, Winbugs, SPSS, GenStat, Mathcad and Maxima. Are these not very widely used? I thought SPSS was - or not in medical areas? I don't mind learning in my own time for a job, but since all my free time is spent on the degree (I am a part-time learner) I don't really have room to spare right now. Perhaps over the summers.

Thanks again and thankyou for the offer of a PM.

Hi there. As far as suggestions regarding suggested jobs, I can't quite think of any that would necessarily put your foot in the door besides those that would some form of computer programming or data analysis type of role. I know that there are low-level data analyst-type positions that could open things up (I know these are available in the Canada and the US, but I'm not sure about the UK). Really, any experience that shows you that you are capable of meeting timelines and show a willingness to learn and be responsible can't hurt you in the least. I would definitely recommend that you pursue the Msc in statistics that you mentioned that is offered through distance learning, and shadow the consulting unit.

Now as far as background in biology -- in my experience, a background in math, statistics or computing is all that is necessary, since any knowledge of biology is something you can pick up while working. The only qualification to that is that there may be specific applied statistics courses that may be more applied to biological problems (e.g. survival analysis, longitudinal data analysis, typically offered in a Msc program) and those I would recommend you take.

On your final point, Minitab has generally been restricted in an educational setting to teach data analysis and is not widely used in industry. WinBUGS is used exclusively for running the Gibbs sampler for Bayesian methods (you'll learn more about Bayesian methods in senior level statistics courses) and is primarily used in an academic research setting, although a few biotech companies have been known to use them. SPSS is most often used in marketing and social science (e.g. survey analysis) and some teaching hospitals also use them as well. I don't know much about GenStat, Mathcad or Maxima.

I would definitely suggest that you learn to program in SAS -- see if you can get a discounted student license through the Open University. R is recommended as well, although it is not as widely used in industry; the one big advantage of R is that it is free, like Python.

I hope these answer at least some of your questions.

 

[3.141592]

Hi again,

If I can afford to leave my undergraduate early and get onto the MSc I will (there is an access course for semi-numerate graduates but I have neither the money nor background for it at this point). I also see they use R.

Thanks for getting back. I have read in many places that there is a fair amount of leeway to talk with the experts on any project/in any role to get the required knowledge. Good to know this is possible in health-related areas too.

That's unfortunate that Minitab and WinBUGS are not used widely in industry but hopefully one reason they were chosen is as they are considered good for giving some understanding of the role of computers in statistical work anyway. I must say I have not really seen adverts requiring either of these. I have seen SPSS in some adverts so that is a good sign I suppose. I don't know anything about GenStat other than that it is used in the final-year module on linear models, but Mathcad and Maxima are, as I understand it, computer algebra packages (they are not on the stats/probability courses).

Cheers for the tip. I better spend some summer time learning SAS or R, depending on price.

Thanks a lot; you've been very helpful!

 

[MarneMath]

Thankyou MarneMath for such a detailed reply, it's very helpful.

I don't know if there is anything meaty in my course on experimental design (but I am hoping so!) but I know there is a chunk on medical stats which includes clinical trials so perhaps it comes up there.

I don't know what 'measure theory' is right now, so I better look that up! I am sure you are right about looking constantly for real world uses (well, uses I am interested in, since it's all useful some way or another); whenever I get stuck and frustrated in mathematics it's always ten times worse if I can't see a relevant-to-my-interests use for what I'm battling to understand.

I read that high-level statistics and probability allow one (to some extent) to flit between industries much more than many other jobs might. Since you began doing statistical work in finance and are now in epidemiology, I wonder if that speaks to your experience? Actually, epidemiology interests me a lot but it looks like one needs higher degrees to break into that field(?).

Your comment reminds me of that quote (Fisher) about a statistician conducting a post mortem of an experiment - perhaps being able to say what it died of! I do like the idea of being a 'go to' guy in a team actually, so long as it is something that requires skill, rather than from laziness i.e. being the 'go to' guy for fetching more paper for the printer... (been there).

Is it reasonably common, if you can say, to be involved with experiments or trials or let's just say a project, from conception all the way through to recommendations based on results? I think I would really enjoy that, rather than getting something on my desk with no idea how it got there, tinkering a bit, then shipping it off with no clue what becomes of it or my efforts. Quite alienating I have found. Also, I don't want a job where I am purely a technical guy, spending all day shuffling symbols on paper or screen. Would you say there is much room for lateral thinking or creativity when figuring out how to collect the data a project requires?

Your description of having a rep as a solid player with stats skills, who is involved in projects from start to end, sounds just wonderful.

Thanks a lot.

Well, since you don't know what measure theory is, I'll make a point to say not to avoid any real analysis course, even if it is not required by your school. However, if you don't plan to go to graduate school, then you can ignore this recommendation.

I'm unsure you have taken any probability or stats course, but if not, when you take your first, always ask yourself what everything really means. What does the number mean to you? What would it mean to someone else? Solving the problem becomes simple enough after practice, the real challenge is looking at the answer and translating it into terms a non-mathematician can understand.

Regarding the type of degree needed for epidemiology. There's jobs at every level. If you want to be very involved in the project and advise the project heads, then you'll need a masters with relevant experience but more than likely a PhD. I only have a masters, so I thus I run models and grab data. I mostly focus on verify and validating models and reviewing results. The people who work under me with only Bs program a lot of the code for me and organize data for me. In turn, I write reports and give them to my boss, and he rewrites it and gives another report. I'm not sure if this is common, or if this is just how government work looks like (from my experience in the military, government work is a series of handing papers to different levels of people.) So with all that said, I've found that you can get a job working within the field of your choice with any level of degree, but the more you have the more 'creative' the work becomes.

*I also feel the need to say that I wasn't originally hired to do my current job. I was hired to do data mining on something completely different, but I wiggled my way into my my current job by way of cocktail parties and a common hatred with my boss towards Faulkner. So with that said, it's very important to form bonds and connections. You never know when they'll be helpful.

 

[3.141592]

Apologies if this is a naïve question but is real analysis the same thing as analysis? In my second year, as well as the stats pathway, I have to pick either the pure or the applied maths pathway. The pure pathway includes analysis, stated as the foundation of calculus. Does real analysis also get referred to as just 'analysis'?

Although not for a long time, I plan eventually to go to graduate school for a statistics MSc, possibly medical stats; in line with your comment about creativity, I read everywhere that an MSc is really the minimum needed to progress as a statistician.

I'm taking my first proper statistics and probability course now. Thanks for the tip. I see what you mean: at this introductory level the computations are relatively easy. Plug and chug. Picking which model to use, or trying to figure out how the value that gets spat out relates to the original problem - and putting this on paper in simple English, I find not nearly so easy.

Report writing comes up later in my course but I have never written a scientific report before. I only ever did arts essays at university before. I read 'good report writing skills' in a fair amount of statistical job adverts. Do you enjoy this part? Is it hard?

The more I read the more I read about the importance of networking. Point taken.

Thanks again.

 

[MarneMath]

Real analysis is often just called analysis or even just advance calculus to some extent.

As for writing technical reports. I should warn you that I have a second major in English with a focus on technical writing. There is an aspect of technical writing that I am quite fond of. I enjoy designing layouts and the challenge involved with breaking down technical information to a level accessible to everyone else on the job. However, I suspect I am an exception to the rule. I also see a lot of people write terrible reports that leave people clueless. Writing good reports is a skill that is learned and one that needs practice.

There's also the other side of the coin, that's writing a technical report for a technical group of people. I'm actually not so good at doing this. I tend too much time explaining what most people would consider obvious statements and perhaps not enough time connecting the overall arguments. Thus I rarely write an official report in this regards, only minor writes up for my bosses.