Building skills through an online business?

[AaronA]

I'd like to move into an online business doing data analysis (primarily surveys, and I'd like to focus on politics). It's a new field to me, but it's something I'm interested in. I'd like to start out small, and see if I can learn and make money at the same time. How do I start out, though? I was thinking an ebook that explains methods of online surveys, and some example analysis done in a blog would be a good start. I could also add affiliate links to Amazon, or to software packages. Any thoughts?

Aaron

[Brutha]

Who do you want to sell to? What your background in statistics?

[Curtis2011]

Find a demand and satisfy it.

[AaronA]

Originally Posted by Brutha

Who do you want to sell to? What your background in statistics?

Ultimately, I'd like to be acting as a political consultant to politicians and people running for office. My stats background are some undergrad and graduate classes (though these were more focused on engineering and business applications), and some training classes. I'd like to go back for a masters in applied stats or mathematical modeling. I can see broader uses in analyzing surveys (particularly written response surveys, which I think is underutilized) in online marketing for small to mid-size companies that don't have in-house departments.

[iDreamCatcher]

Data analysis is a big business. Or data mining rather.
See if you can offer unique perspective, unique view that helps to answer questions that no one can.

Doing what others do (such as amazon affiliate links) is waste of time.

[Brutha]

Ultimately, I'd like to be acting as a political consultant to politicians and people running for office.

I have a hard time imagining the a politician who wants to hire a pollster to hire someone online instead of a pollster that he can meet face to face.
If that's what you want to do I would rather recommend getting a job at some agency that does political consulting.

I can see broader uses in analyzing surveys (particularly written response surveys, which I think is underutilized) in online marketing for small to mid-size companies that don't have in-house departments.

Usability people can run statistics, doing Adwords optimization would be another area where statistics is needed.

I'd like to go back for a masters in applied stats or mathematical modeling.

Advanced mathematical models are used in academia but when it comes to the needs of small/mid-size companies you don't need fancy mathematical models.

A statistics degree also runs the risk that you start believe too much in a Gaussian world because the textbook problems are all Gaussian.

[iDreamCatcher]

I remember interesting example about the guy who made a survey in a medical field. He just sent marketing/survey email to his list - something like: "which medication helps you more against ....[some condition]".

He got some replies and published it on his website. Due to the nature of the information in his survey results page - it was picked up by About.com and they added ~2000 backlinks to his survey page and that overnight established him as authority on the subject.

Gleb

[AaronA]

@Brutha Actually, I think the academic world (at least the folks I've been reading) have been going in a more Bayesian direction. And the political angle would be something I'd need to work up to. There are a lot of data mining techniques that I think have a lot of potential, and aren't in the mainstream yet.

@iDreamCatcher Thanks for that example. I really think that I could get some positive exposure by posting things like that on-line. Plus, I want to do this in order to explore a new field that I enjoy, and doing work that interests me vs. something that a boss assigns me to do really appeals to me.

Aaron

[Brutha]

@Brutha Actually, I think the academic world (at least the folks I've been reading) have been going in a more Bayesian direction.

You still have the problem that you work with data in university courses that are designed for the technique that you are working with while real life data often isn't.

The difficulty with real life statistics isn't getting answers but understanding when the technique you are using produces wrong answers.

And the political angle would be something I'd need to work up to. There are a lot of data mining techniques that I think have a lot of potential, and aren't in the mainstream yet.

In business you don't get extra points for using advanced statistical techniques while you can impress peer reviewers with them in academia.
In business you want techniques whose limitations are well understood.
That's especially true in a field like politics where you don't get good feedback that tells you when your models are wrong.

If your statistics are wrong in academia and fancy enough to get past peer review you have a published paper.
If your statistics are wrong in business than you get bad business results.