Posts Tagged ‘Data’

Watch out, that ice cream might be a murderer

Tuesday, February 16th, 2010

I just learnt a very important lesson that will serve as a mantra during my statistical analysis for my thesis.

“Ice cream does not make people kill each other!”

A researcher found a correlation between ice cream sales and homicides, which in simple terms would mean that as soon as people start eating ice cream they’ll go on a killing spree, or that after a killer struck, s/he goes on a ice cream binge.

Luckily, this not true. It is a classic example of partial correlation. What you need to find is the third variable that is missing from the equation. In the example fom above – weather. When do people eat ice cream? And when do people spend a lot of their time outside? Summertime. A does not cause B, or the opposite, you need to find variable C.

I will nurture this little lesson dearly.

Netflix – They will know what you like, without asking you.

Sunday, October 18th, 2009

Reed Hastings, CEO of Netflix, is interviewed in number 18 of the European edition of Forbes. The article tells us that Netflix have, using a cash prize of $1,000,000, made findings concerning movie grading and peoples’ movie watching habits. They can, more accurately than their competitors, suggest movies to their customers that are similar to those that their customers have graded. This means that a person who likes Action movie A and Chick Flick B will probably like Horror movie C.

To describe the problem, Mr. Hastings says that:

The trick is finding problems in your business that you can package cleanly, where you can provide a sample data set and a very clear evaluation metric.

What the cash prize of $1,000,000 gave the company was a discrete math problem that will predict future movie ratings using past movie ratings. A must add to the prize itself, is that it was given to a group of people, who many of them, had never meet. Someone started out with a solution, posted it on the Internet, got help from someone else. So it continued until they had solved it, and they all got their share of the dough.

Reed also gives a short glimpse at the future and says that they are now working on is to determine the movies that will appeal to a customer who don’t rate movies online, at all.

Imagine that this was your company. What would you use this form of information to?

Could you use it to test whether or not a certain product would be a fail or not, or maybe which demographic it would be best fit for.

How about the way the solution was thought up, using several different people, who might not even live in the same country, know each other, or might not even meet, to help you with your development.