A while ago I wrote this:
Also, a 30-day moving average should be insensitive to a dip in traffic on the weekends, right? Our 30-day moving average at Sitemeter dips every 7 days and we can't figure out why. I e-mailed Sitemeter with this question and the response was something like: the 30-day moving average is measured using a methodology that is so complex that we can not reveal it to Sitemeter users.
And, I got a polite comment from David Smith at Sitemeter explaining how simply a simple average is calculated. So, what does a dumb guy do when he can't figure something out? I took the sitemeter data before the x-idays and what to my wondering eye should appear? The 30-day moving average dips every 7 days! Click on the image above for our first 6 months day.
I've also figured out a few other sitemeter puzzles that bugged me that shouldn't have: (1) the "average per day" is the 7-day average, (2) the average per day will rise on a slow day if the slow day replaces a slower day in the average, and (3) the 7 day moving average graph is one day behind the average per day so the numbers don't match.
Update: David Smith at Sitemeter suggests comparing the 28-day and 30-day moving average (click the thumbnail). As he suggests, the 28-day moving average smooths out the "weekend effect" that made me so ornery.
