5
$\begingroup$

I am graduating with a bachelor's degree in applied math and I will pursue a master's degree in statistics this fall.

There are many specialized fields in applied statistics. I realize I may be more interested in becoming a data analyst, working with big data in market research or IT, but I'm not sure what kinds of skills are needed and what are the core courses I should take to become a data analyst.

I've learned some MatLab, R, SPSS, C, ruby, can't say I'm an expert on these. I'm learning SAS, because SAS seems to be what every stats master should know. (Is this really true?)

I'm a little confused right now and need some specific advice about what skills I should intensively learn and what courses I should take.

TIA.

$\endgroup$
8
$\begingroup$

Aside from the technical skills, like R or SAS, SQL will be important, and a few other higher level skills, including:

Data Manipulation: To be able to analyse data, you will frequently have to spend some considerable time acquiring the data and manipulating it into a form which can be analysed. Many statisticians will tell you that most of their time on a given project will be spent manipulating the data - so it is important to be good at this!

Understanding: Many people vastly under estimate the amount of time that is required just to understand a complex dataset. In bygone days one had to serve time apprenticed to a master crafts man, with a dataset you have to spend time looking at the various facets of the data and understanding it's dimensions and missing data and talking to people to try to understand the data. Again you will spend considerable time doing this, it requires practice to build up this skill!

Visualization: Going hand in hand with the above is visualization. Knowing how to plot data to help gain understanding is important. Later when you want to show somebody else what you have found in the dataset, a carefully created picture says a thousand words.

Requirements: One of the hardest to learn is requirements gathering. Your customer will frequently not be clear what they want in their own head, and even less clear in what they say to you.

$\endgroup$
  • 1
    $\begingroup$ (+1) Learn to work with data. A very nice supplement is Problem Solving: A statistician's guide by Chatfield $\endgroup$ – Tim Jun 9 '12 at 17:14
  • 1
    $\begingroup$ Regarding visualization, Edward Tufte has an excellent collection of books he has authored on the topic. The Visual Display of Quantitative Information, Envisioning Information and Visual Explanations: Images and Quantities, Evidence and Narrative are all great reads. Protovis and D3 are also both excellent visualization tools. $\endgroup$ – mwjohnson Oct 3 '13 at 0:59
2
$\begingroup$

SAS is important in the pharmaceutical industry but not necessarily in other disciplines. in business and marketing time series analysis and survey sampling are particularly important. Yes big pharma and business use SAS a lot but it is expensive and makes more economical sense with multiple users that you would find in big companies. in the social sciences SPSS is more commonly used but is also expensive. R is free and many advanced statistical procedures can be found in the CRAN libraries.

$\endgroup$
  • $\begingroup$ thank you. should I just focus on R? I heard big companies do use SAS a lot, is it just a misunderstanding? $\endgroup$ – Ivy Lee Jun 9 '12 at 12:45
  • 1
    $\begingroup$ @IvyLee: Google R vs SAS to get a feel of the existing discussions on R and SAS. Nothing wrong with learning both and using whichever you prefer. However, for learning statistics, I would recommend focusing on R. It is open source, and you can easily view the source code to learn how statistical algorithms are being implemented. $\endgroup$ – jthetzel Jun 9 '12 at 13:07
  • 2
    $\begingroup$ There is also an age divide in some industries, e.g. some younger actuaries prefer R, older actuaries have been using SAS for years and prefer not to change. $\endgroup$ – Sean Jun 9 '12 at 16:10
  • $\begingroup$ I understand the advantages of R. I've been a Linux desktop user for more than 5 yrs and I embrace open source and the idea of free software. The only reason I'm considering learning SAS is that SAS appears quite frequently on job descriptions, I'm afraid that if I only know a bit of SAS and lots of R companies would not favor me, after all they spent a lot of money on SAS software they'd expect employees to use it. I can only learn SAS with the help of school's SAS server since Linux is my only OS. $\endgroup$ – Ivy Lee Jun 10 '12 at 5:31
  • $\begingroup$ @IvyLee Your idea about SAS makes a lot of sense. I would encourage you to learn it. Given that you know how to program in R, SAS should not be difficult for you. $\endgroup$ – Michael R. Chernick Jun 10 '12 at 11:48
2
$\begingroup$

I would recommend that you take some microeconometrics courses. Business analysts spend an inordinate amounts of time patching up analyses that don't make much sense once you've done this. One common example is running regressions of revenue or profit on price and believing that these estimate some sort of demand elasticity. A lot of media mix modeling falls into this. Understanding endogeneity and counterfactuals goes a long way in the business world.

I would also make sure I know how to clean data, since most of your time will be spent doing exactly that. This spans hard skills like SQL, and perhaps some Hadoop/Pig/Hive, but also being able to do quick gut checks. I call these softer skills number-sense. These things are rarely taught in master's programs, where the data you see comes packaged with ribbons and bows.

$\endgroup$
1
$\begingroup$

In addition to the many great suggestions, I'd add that you should try to learn some soft/people/analysis skills and perhaps at least a bit about one field where you might want to apply your skills as well.

In the real world, no one hands you a clean set of well-documented data with a precise question to answer. You will need to massage the data, understand the data, pull expertise out of people, understand the data-owner's business, beg for data, present interim results, etc. You may often end up using embarrassingly-unsophisticated techniques. (I hope you don't expect labeled data.)

Don't neglect statistics and sophisticated techniques. But don't think that you'll spend a significant portion of your on-the-job time doing and thinking about them.

$\endgroup$
0
$\begingroup$

I recommend reading these two before doing anything. Hope they can shape your mind:

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.