# What is it that a statistician does?

When I tell my non-statistical friends I am a grad student pursuing a PhD in statistics they naturally say "oh so you want to be a professor?". I tell them no I actually plan to work in industry. Then they respond with, "and do what?". I have not found a good response to this question. I want to give them a range of interesting problems that statisticians work on, but my response is usually muddled or is too technical. I think most people picture us working in excel and calculating means and standard deviations.

What is a good response to this question that is brief and interesting?

Thanks! (not sure what a good existing tag is for this question)

• I threw on the careers tag. I think that fits somewhat. Also thinking this should be CW. Dec 14, 2010 at 3:48
• "Statisticians are people who try to come up with meaningful conclusions from messy data." - unknown Dec 14, 2010 at 8:54
• I think it is two different questions. The first is the one you asked, the second is what to tell your friends. It depends on friends really. I think saying that you will do consulting work, where you consult firms on solving problems with their data, will be enough for the majority of your friends. At least for me it works :) Dec 14, 2010 at 9:19

The area I am most interested in is the realm of biostatistics. Statistics can be used in this regard to do anything from summarize the results of a drug trial, determining whether Prozac really is more effective than the placebo sugar pill, to tumor detection in cancer patients. Please check out this presentation I found:

What is Biostatistics?

Remember, a statistician is a function that maps a set of data to a set of decisions.

• Your last sentence - brilliant! Dec 14, 2010 at 7:02
• This link seems broken by now. Any update?
– chl
Oct 16, 2011 at 10:28
• @chl: It should be fixed. Oct 17, 2011 at 15:46

A statistician is a numerical detective, uncovering the stories hidden in a mass of data.

A statistician performs the following generic steps:

1. Has a goal- test a theory, generate a prediction etc

2. Obtains data (via surveys, experiments, thrid-parties etc) that is consistent with the above goal.

3. Develops a statistical model that says roughly: Variable of interest = function of covariates + random error.

4. Estimate the impact of covariates on the variable of interest using various tools from math/stats.

5. Use the resulting estimates to evaluate theory/generate predictions.

You can contextualize the above generic steps to various settings. The context can be changed depending on the audience. For example:

• With doctors: discuss how you would assess the effectiveness of drugs

• With a more general audience discuss how you would assess the winner in an upcoming election

• With business professionals discuss how you would assess the relative impact of advertising designs etc.

I'm not a statistician, but a lot of my work involves statistics, and I work in health care.

The two things that I spend most of my time doing are:

a) examining the sizes of effects and trends and seeing if they are "real" b) presenting very large datasets in a simple way so that managers and users of our services can understand them- usually in the form of graphs.

Having said that, I have NEVER successfully explained my job to anyone at a party, so I think I'm in the same boat as you! I love it when people say "that sounds interesting"- because it doesn't, not the way I tell it!

The TV show Numb3rs is useful as many people have seen it. I tell them that I'm like the guys on Numb3rs except I deal with solving business problems rather than crimes. (Substitute "business problems" for whatever field you work in.) That usually gets the response "Wow, cool!" which is better than what I used to get.

• There are snippets of Numb3rs that explain some statistical concepts extraordinarily well. In 60 seconds or less, Charlie explains some concepts that would take most teachers 10X as long. Of course, there are bloopers, but the good material is quite good. Oct 15, 2011 at 14:17

My attempt at a simple answer that is both applicable across sub-domains and understandable (in gist) to the lay-person: When science develops theories about the world, these theories are compared to real-world data. The role of the statistician is to assess how well one or more competing theories account for the data. This is achieved with mathematics that let the statistician quantify their uncertainty about the conclusions they draw from the comparison of theory to data.

A statistician tells you what conclusions can be reached from a data set and, maybe even more important, what conclusions cannot be reached.

I work as a statistician for a business intelligence team at an online retailer. At my job, I'm frequently building models to predict various things (like response rates to catalogs, open rates of emails, etc.). I also help marketing set up A/B tests (like, does email A do better than email B)? Sounds like a simple problem, but this can be pretty complex when you start to do some power calculations for sample sizes. Additionally, decision-makers in businesses always want to know if profile A differs from profile B. For example, did we sell more black shoes than brown shoes this year? Did we convert more lapsed customers with this catalog? Did the marketing channels people converted on differ this year from the ones used last year? These questions are (probably) simple to answer on the surface, but it requires statistics to understand if these answers are significant.