Corrupt countries were more likely to support OOXML
|
| Author | Content |
|---|---|
| henke54 Sep 05, 2007 9:29 PM EDT |
Quoting:We studied the relation between the corruption level and voting behaviours of the countries. We found that more corrupted the country is, the more likely it was to vote for the unreserved acceptance of the OOXML standard proposal.http://www.effi.org/blog/kai-2007-09-05.en.html ;-P |
| jacog Sep 05, 2007 11:11 PM EDT |
South Africa is pretty corrupt though (I know, I live here) and they voted no. Lucked out there I guess. :) |
| dinotrac Sep 06, 2007 2:56 AM EDT |
Don't take too much out of that study. The numbers are small, and the perceived corruption numbers almost certainly covary with about a million other things. For that matter, the very notion of perceived corruption is a tough nut. For example, is it a few western countries looking down their noses at the rest of the world? Do people in each country provide the perception? If so, does it mean the same thing from country to country? And so on., |
| NoDough Sep 06, 2007 6:04 AM EDT |
Corruption is a human condition, not a geographical condition. It's true that certain governments create an environment where the practice of corruption is less discouraged, but corruption itself is a human trait that can be found in any population of any country. The index used in the article is the Corruption Perceptions Index. Great, now we're measuring perceptions as a matter of fact. Never mind that perception=opinion. And never mind that people are very likely to lie about their perception in order to bias the study. And never mind that perceptions are largely a construct of the media. In short, I have zero confidence that such a study proves anything. |
| mvermeer Sep 06, 2007 7:12 AM EDT |
When reading this I was thinking "Why doesn't he use the numbers from Transparency International?" ...but then I saw that this is the authoritative TI survey that's being used. Yes, it's called "perceptions", but that means the perceptions of interviewed experts, not the people at large, as is documented on the TI web site. This is the most reputable survey of its kind. It is also true that correlation doesn't prove causation. If A correlates with B, that may mean that A caused B, B caused A, or both have some common cause C. It is e.g. clear that corruption correlates with poverty, due to various reasons: poor people are more easily tempted, or (probably more realistic) poor countries are more likely to lack the democratic controls to bring corrupt officials to task. Still, the methodology in this article is valid. You formulate a hypothesis: "corruption -> OOXML vote", then you design a statistical test, and you get an outcome: hypothesis accepted/rejected. This hypothesis makes sense, it is based on a well-understood mechanism. What others are on offer? |
| dinotrac Sep 06, 2007 7:24 AM EDT |
>What others are on offer? The lack of good information does not compel you to accept meaningless stuff because it was done well and looks a lot like studies that actually do have sufficient appropriate data to support their findings. More to the point, one must always apply a high standard of scrutiny to any study that tosses the trappings of methodology casually over something that uses the output of an advocacy group for its measuring stick. |
| mvermeer Sep 06, 2007 10:51 AM EDT |
> More to the point, one must always apply a high standard of
> scrutiny to any study that tosses the trappings of methodology
> casually over something that uses the output of an advocacy
> group for its measuring stick. Huh? TI is an advocacy group -- if that -- against corruption, not against Microsoft. And why don't you trust TI by the way? Do you like corruption? And even if they were dishonest, they lack the power to manipulate the laws of statistics. The correlation stands. Even if the input on corruption levels had been produced by drawing the horoscopes of the respective countries' heads of state... where there is significant correlation, there is something to explain. What that explanation is, is something statistics cannot answer. |
| dinotrac Sep 06, 2007 11:02 AM EDT |
>where there is significant correlation, there is something to explain Where there is a significant correlation, there are numbers that move together. |
| Bob_Robertson Sep 06, 2007 11:13 AM EDT |
> Where there is a significant correlation... ...there is a significant tendency to see causation where none exists. |
| dinotrac Sep 06, 2007 11:15 AM EDT |
>...there is a significant tendency to see causation where none exists. Sort of like... Where there's smoke there's fire or possibly, a car that needs new rings. |
| mvermeer Sep 06, 2007 11:41 AM EDT |
Actually you can put numbers to that belief: for the national body vote,
the probability of the correlation happening just by chance, without any
real cause, is 1.9%. Events with 1.9% probability do happen now and
then. Whether you want to believe that is between you and your
conscience ;-) Actually Dean's reservations (small numbers, and mere statistics don't explain anything) are stated in the article. Sometimes it's a good idea to read before commenting. |
| dinotrac Sep 06, 2007 11:50 AM EDT |
>Actually you can put numbers to that belief: for the national body vote,
the probability of the correlation happening just by chance, without any
real cause, is 1.9%. Ummm... You'll have to explain that one to me. I haven't done the calculations, but I am reminded of Richard Feynman's fondness for grabbing a license plate number from the parking lot and asking something along the lines of, "Can you believe it? I saw license number xyz1234 in the parking lot today? Absolutely amazing! What are the chances of that?" Students would go to work trying to figure out the odds, but, after a while, he would ask them what they were trying to calculate. There is no point trying to calculate the odds of seeing a particular plate unless he had gotten up that morning and written the number down. The real question was "What are the chance you'll see a license plate you don't know?" Those are pretty high. Throwing out probabilities without explaining the calculations or why you think some particular random model should apply is little more than mental masturbation. |
| Bob_Robertson Sep 06, 2007 11:55 AM EDT |
Don't be too rough on him Dino. Everyone knows that 73.4% of statistics are made up on the spot. |
| NoDough Sep 06, 2007 12:01 PM EDT |
Quoting:Don't be too rough on him Dino. Everyone knows that 73.4% of statistics are made up on the spot.Actually, it's 86.3% |
| dinotrac Sep 06, 2007 12:44 PM EDT |
>Actually, it's 86.3% +- 1.567%, approximately. |
| azerthoth Sep 06, 2007 4:54 PM EDT |
Statistical numbers are also subject to one other problem, a human has to present them. Using the exact same numbers you can present the argument as either a positive result or a negative result and either will stand. Many times when people refer to "finessing" the numbers they are actually discussing how to massage the numbers to mean what they want it to mean or think it should mean. |
Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]
Becoming a member of LXer is easy and free. Join Us!