Sophie - November 14th, 2014
my journal
September 2016

Date: 2014-11-14 17:53
Security: Public
Subject: An interesting statistical anomaly

So I discovered an interesting anomaly in the Dreamwidth stats recently.

One of the things I do is keep historical daily copies of the raw stats file for Dreamwidth, which is updated daily. I've been collecting this file automatically every day since September 2010, and today I decided to take a look at today's file.

One thing I noticed was that the number of accounts with a birth date listed that would put them at 5 years old seemed to be really quite high compared to its surrounding data. While the number of accounts "aged" 6-12 totalled 107, the number of "5-year-olds" alone totalled 259.

It's worth noting at this point that you can't create an account if you put in a birth date that would indicate that you're below 13 years old, because Dreamwidth doesn't want to have to deal with COPPA. In other words, to have an account listed as any age between 5-12, you'd need to have created the account with an earlier birth date and then adjusted it after creation to be the birthdate you wanted it to be.

I decided to take a look at the historical copies of the stats file. The biggest part of the rise seemed to have taken place mostly in April-May 2014. At this point I plotted a graph to show the statistical anomaly, which internally I felt very conflicted about doing. (Without knowing why the graph was plotted or seeing the data, it sounds very creepy to plot a graph plotting the number of accounts with a birth date indicating they were 5-12 years old over time.)

My first thought was that these must be bots or RPers. I shared it with Dreamwidth's IRC channel and we all pondered over it, before [personal profile] pauamma realised what must be happening.

It turned out that what we were seeing was basically a replay - much diminished, mind you - of the growth Dreamwidth experienced in open beta when it started in April 2009 - 5 years ago! Some of the accounts created then must have set their "birth date" to be the date of the journal's creation - an account birth date rather than the user's birth date. We hadn't noticed before now because the stats file only goes down to 5 years old, and even though there were accounts in closed beta, the figure wasn't high enough to particularly notice before now.

What I'm saying is, we basically missed Dreamwidth's fifth birthday, if we take "birth" to mean the start of open beta. Dreamwidth is officially five years old!

4 Comments | Post A Comment | Add to Memories | Tell Someone | Link