European governments have long complained about their dependence on Microsoft's software, but their rhetoric has not turned into a mass migration away from Windows. During the past few years, Europe's elected officials have made a lot of noise about ambitious projects to switch to open source software, including big migrations of government PCs in France, Germany, Spain and Norway. Yet the actual migrations have been negligible. More than 95 percent of all PCs used by European government workers still run on Windows, according to the market research firm IDC.
To view parent comment, click here.
To read all comments associated with this story, please click here.
Member since:2006-06-01
You send forms to each member
Theres the problem. You don't understand weblogs. No one is sending anyone a form.
Think of it this way. A person stands on a street corner and counts the type of car that goes by. Day after day after day.
The bias would be geographic location. If the corner was in the US, then more american cars would show up. Etc etc.
None of those people in the car were sent a form. They just drove by.
Now, what happens when someone stands on a 75,000 street corners in 100 countries and record the make and model of each car.
Thats a HUGE sample. No forms. People going about their day to day business.
And then you take a random sample of 20,000 of the recorded make and model, each from 100 countries.
Day after day after day.
Thats what onestat does.
The ONLY bias you can say for sure is that it discriminates against people who don't own or use a computer.
OS and browser figures on Onestat, corroborated on 10's of thousands of other "street corners", show Linux with about .36% of the market.
Live with it.
Member since:2005-07-02
Theres the problem. You don't understand weblogs. No one is sending anyone a form.
Man you are dense. It was an *analogy*.
I do understand weblogs. I've administered them. I've analyzed them. I've used software similar to the one OneStat uses.
Now, what happens when someone stands on a 75,000 street corners in 100 countries and record the make and model of each car.
Yeah, except the 75,000 street corners are not selected at random. Therein lies the rub.
The ONLY bias you can say for sure is that it discriminates against people who don't own or use a computer.
No, it discriminates against people who don't visit OneStat's customers' web sites.
If it was 75,000 random web sites, then that would mean a better sample (although it *still* wouldn't be an accurate measure of market share, because web stats cannot give the number of individual users visiting the site). But it's not 75,000 random sites, it's 75,000 user-selected sites.
OS and browser figures on Onestat, corroborated on 10's of thousands of other "street corners", show Linux with about .36% of the market.
Web stats cannot be used to determine market share, regardless of sample size, because web stats cannot give an individual user count. *Anyone* who's ever analyzed a web log knows that, and the two research papers I linked to corroborate this information.
Web stats cannot be used to determine market share. Live with it.
Member since:
2005-07-02
It is statistically significant if it involves 75,000 collection points day after day after week after week and the numbers all match.
Wrong. We cannot deduct that a sample is biased through consistency.
Let me give you an example. Let's say you want to have a good idea of the results in the upcoming election and so you conduct a poll. However, instead of using a true random sample method, you decide to go through your business contacts from the State Association of Chambers of Commerce. Let's say that the Association has a membership of 10,000. You send forms to each member and ask them to fill in the poll questions. You get back the forms and find out that Republicans are getting 67% of the vote. The high number of people responding means your margin of error is very small.
However, you forgot to take into account that businessmen big and small, who make up the bulk of chambers of commerces throughout the country, generally tend to lean more towards the Republicans than the democrats. Therefore, your results are skewed and do not pain a reliable portrait of the overall political picture.
It's the same here. OneStats clients are *not* random, they are self-selected, and self-selected samples will *not* provide an accurate picture.
That's besides the point, though, because web stats cannot provide the number of individual users anyway, and so they are useless to provide market share. Of course, OneStat is not going to admit this outright, but they do talk of "usage" instead of market share, which is telling.