Fun with statistics: estimating blog readership (a do-it-yourself recipe)
Posted by Tom Moertel Thu, 23 Aug 2007 01:34:00 GMT
As everybody knows, statistics is fun. Is there anything cooler than crushing a heap of seemingly uninteresting numbers into gleaming jewels of meaning? Of course not! Models, data-visualization plots, and fat data sets are way cool. So, let’s find an excuse to play with them.
Here’s an excuse – I mean, an important and highly relevant question that many of us share: How many people actually read our blogs? To answer the question, we will need to use statistics, data, and cool plots. Further, if you’ve got the raw data for your blog, you can follow along with your own analysis. Even more fun!
We’ll start with a simple inspection of common web-log data, using command-line tools. After developing a rough understanding of what useful information we can extract, we’ll analyze the raw data using a series of successively more sophisticated techniques. In the end, we will derive a simple formula for estimating readership from easily obtainable data.
Sound good? Then let’s get rocking.
But first, a preemptive strike on would-be poo-pooers: I know all about FeedBurner. I know they will track my blog’s subscribers and use their mystical powers to infer the number of “real” subscribers I have. I know it’s all so easy. But easy isn’t the point. I want to understand what’s going on. Just taking somebody’s word for it isn’t nearly as satisfying as figuring it out yourself – nor as fun.
OK. For real this time, let’s get rocking.
Read more...
readers