Wednesday, May 23, 2007

Lies, damn lies, and statistics

There are three kinds of lies: lies, damned lies, and statistics.
Mark Twain attribution to Benjamin Disraeli
I don't normally head over to Groklaw, but I did recently based on some other links and came across an article about a study conducted by Microsoft on how developers don't want the GPL, version 3, to "police patents." If you dig a little deeper into the study, the following facts about the study are exposed:
  1. It was conducted by email.
  2. 354 emails were sent out between Feb 28 and April 4 2007.
  3. 332 reached their destination.
  4. 34, or slightly more than 10%, responded.
Now, PJ chose a particularly inflammatory (some might even go so far as to say trolling) title: "Only 11% of OS Targeted Programmers Willing to Help MS-Funded Study." And that's because Microsoft did indeed fund yet another study seeking to bolster their position with regards to the GPLv3. I don't see it quite that way. I see that only 10% of the targeted group (according to my simple math skills) bothered to respond (frankly, for reasons unknown). And those 34 bear the heavy responsibility of representing the rest of us who either refused to respond or who were never even contacted. Great responsibility indeed.

Do I feel pity that Microsoft's efforts were misconstrued yet again? Hell, no. Microsoft has a long history of lying, cheating, and outright theft in the accomplishment of its singular goal of total domination in all markets it wants to play in. Microsoft went looking yet again to buy more support for their position with regard to the GPLv3. And Microsoft got sloppy in deciding to run with this particular study. Every Microsoft 'study' been pretty dodgy to my eyes, but this one seems the worst by far. This particular study is based on only 10% of it's statistically significant sample group. And I'm assuming, based on my ancient statistics classes from some 30 years ago, that 332 (not 34) is the minimum population size required for the study to have any relevance. So if you only got a 10% response from your relevant group, how can you really draw any conclusions?

The one conclusion that I can personally draw is that those that did respond are the ones who have an issue with GPLv3. Which means that I can interpret the statistics to mean that out of 332 potential respondents 298 (90%) did not have a problem with GPLv3 and so did not bother to respond. I also wonder that when the initial request went out if the potential respondents were told that Microsoft was funding the study. Would more have responded if they knew? I would have just to make sure my opinion was at least recorded. And I believe many others would have as well. But if it went out as a low-key request to respond to these questions, then many would have read it and tossed it as one more thing to try to shove into an already busy schedule. And so it got dropped on the floor.

The PDF of the paper is here. I'm going to read it and see what the paper's authors really said, and read their own words about how they interpreted the results. I don't trust the eWeek article sited by Groklaw, and I've learned not to trust PJ's interpretation either. Just like I've learned to question just about everything Microsoft says.

No comments:

Post a Comment

All comments are checked. Comment SPAM will be blocked and deleted.

Note: Only a member of this blog may post a comment.