Adventures with Google

The internet is such a fabulous tool for research. The number of resources available to you is simply amazing. Pick a topic, and you're likely to find some information about it using the Google search engine.

Recently, I've been looking for some technical papers related to the research project I've chosen to complete my Masters in Statistics. My topic is a method called Iterative Proportional Fitting (IPF)--also known as Raking, RAS, SPREE, and a few other titles.

Here's the process in a nutshell. Imagine that you have a statistic (such as average income) that is given for different groups, such as age and sex. Let's say that at 20 years-old, men make more money than women. If the difference between male and female income changes (i.e., the difference approaches 0) as age increases, then we have a relationship (or interaction) between age and gender. IPF is a very simple method for updating estimates with new information while preserving relationships between different, specified groups.

Most of the resources I have are about 20-30 years old (some are older), so I'd like to find some resources that are a little more recent. (Cue Google to enter stage right. Play triumphant music. Cue loud record scratch noise.) Some of my attempts were less than successful. Here's a list of my searches and results.

1.) Iterative Proportional Fitting

Result: This search initially offered the best results. Unfortunately, most of the hits resulted in documentation on IPF in different statistical packages. There were a few decent articles, but nothing substantial. There was also some lame SAS code. Overall, this search was somewhat disappointing.

2.) IPF
I've often heard the procedure mentioned with just its acronym. Of course, so many things in government are referred to simply by an acronym, but I still thought it was worth a shot.

Result: International Powerlifting Federation. ..., ah, no.

3.) SPREE
This is a term coined in the early 80's by some researchers. I've seen it referred to in other documents, and it could be used as a keyword in a paper, thus increasing the chances of finding a good site.

Result: Unfortunately for my search, Google doesn't do case sensitvie searches. That is, "spree", "SPREE", and "SpReE" are all the same search. Think of all the phrases that include the word "spree", and you'll get an idea as to how wide from the mark I ran. I ended up with over 9,000,000 links, and links that ranged from shopping spree to Spree candy (I had forgotten about those candies. There's a website that sells "old-time" candy, and it sells Spree).

4.) SPREE statistics
I figure that that ought to narrow the search and make it a bit more specific.

Result: The first website discussed the proper way to rank players in online first-person shooters (top choice is the kill-to-death ratio). Another site gave statistics for killing sprees for online players of Castle Wolfenstein. Definitely not what I'm looking for, though it was kind of interesting.

The IPF method will be used with relation to particular statistics like death or unemployment, so I tried those words next.

5.) SPREE statistics death
Result: Some pro-death penalty and pro-gun websites.

6.) SPREE statistics unemployment
Result: Once again, a few useful articles, but nothing substantial. I'm now down to about 87,000 links. Since I started at around 9,000,000, I'm definitely improving. At this point, I'm hoping for a pseudo Google whack (pseudo since it will involve more than 2 words).

7.) SPREE statistics categorical
I decided to use this hit since SPREE is also defined by that fact that it uses categories.

Results: At last some useful material! I found a couple valuable papers with this hit. I got my search down to about 560 sites--not the 2 I need for a Google whack.

At long last, after about 60-90 minutes (I spent too much time on the first-person shooter stats) my adventure ended, and I had 3 additional papers to read through in my attempts to acquire greater knowledge on the subject.

All in all, it wasn't a bad experience. While it took a little searching, I did find what I needed, and some hard work is usually required to get what we need.

My experience could have been much worse. It could have been an interview with Ask Jeeves.