A Mathematical Statistician

Apparently no one knows what a mathematical statistician does. (I'm pretty sure this includes my parents, so don't feel bad if it includes you.) My friend Todd is a neuroradiologist (see his family's blog here) and he recently did a series of posts enlightening people to what it means to have his job. Oh yeah, even this neuroradiologist doesn't even know what I do, so now you can feel really good about you!

Anyway, I won't say much about exactly what I do because I like keeping work and this blog very, very separate, if for no other reason but to protect myself. So forgive me for being vague.

At a minimum, most statisticians' jobs require a bachelor's degree in math or stats. If you end up being a biostatistician, then you likely need a master's in it. Anyway, I have the bachelor's in math (with an emphasis in stats) and a master's in stats. Which on paper, should mean I'm qualified for most jobs. In practice, I feel qualified for none. The second Bachelor's degree in Geography was just because I like maps.

Anyway, I happen to work on one particular survey that goes out every year (think Census but instead of every ten years, mine goes out every year). And it surveys businesses, not individuals. We don't survey all businesses in the United States as there's about seven to eight million of them in existence, so we do something called sampling. This means sending forms to a selected group of businesses that we hope represent the entire population of businesses. After we send out the forms, data start coming back in. Then we enter the data and start analyzing.

But what does that mean for me and what I do? Well, before we can decide which businesses to sample, we need to decide how many units we're going to sample. That process is called allocation. So I work on that too. After we allocate, we get to select the units, or do the actual sampling. Of course, all of this is done with computers, but there is some manual intervention to make sure we're not taking too many units in some industries and too few in others. The data are collected and entered into a computer, but not by me. Once they all come in, we start the analysis. Actually, as a government agency, we don't do much analysis, just a lot of reporting facts. However, because sampling was involved (as opposed to getting data from every single business), we need to do something called significance testing. If we claim that something increased, we need to make sure it did, rather than just the numbers changing due to differences in sampling. So I test statements to make sure they are correct.

Because our survey is annual, I get to repeat these tasks yearly. Every year things change and we have new things that come up, so it's not as monotonous as it sounds. Tweak this here, tweak that there. It's all very exciting.

Sometimes I get to work on other "pop-up" projects as well. I really enjoy the ones where I get to learn new things, especially when it comes to programming in SAS (a statistical package that's super cool). I also get to take courses on stats stuff.

And sometimes I even get to travel. When I do that, I'm usually giving some sort of training to the non-D.C.-based staff working on the survey.

So that's my job. Clearly there's more to it, but I don't want to lose you. Oh yeah, and I get to enjoy a lot of air conditioning.