News Article: drill attacks on the raise in Iraq[views:4554][posts:52]____________________________________ [May 17,2006 5:39pm - PatMeebles ""] Who knows. I'm just saying that if they were more certain of a specific number, they wouldn't have to put a 95% 8,000-194,000 as a result. If they had a higher certainty it was at 100,000, they would've put a 95% 90,000-110,000, and then there wouldn't be any controversy. |
| ___________________________________ [May 17,2006 5:50pm - ShadowSD ""] OK Pat, I just asked a couple statistical questions of my father, who is a mass media communications professor with a Stanford PhD and expertise in statistics, and I can now say for a fact that the article you cited doesn't know what it's talking about, because the methods it criticized are inherent to statistical study: 1. Percentages and raw numbers are two different things. The analogy with Bush getting between 8% and 94% of the vote makes the reader say "well margins of error are only like 4% for most polls, so that means Lancet's study must be garbage!" In fact, the article's reasoning is not only faulty but deceitful in presenting that false analogy. Why? Well, 4% of the number of people who voted in the last election is over 4,000,000 people (which is way higher a raw number than the disparity between the high and low numbers used by Lancet). 2. Your suggestion that Lancet came up with 98,000 after coming up with 8,000 and 194,000 is WRONG, and CANNOT BE TRUE. In all statistical study, the middle number is determined first, and the low and high numbers come afterwards; in otherwords, the low and high are determined based on the middle number, NOT THE OTHER WAY AROUND. 3. The middle number is ALWAYS the most accurate one. Also, the higher the confidence number, the further the low and high numbers MUST BE from the middle. Those two sentences right there do away with the main arguments criticizing Lancet. 4. I asked if the middle number could be "roughly in the middle" of the high and low (as the article says). ABSOLUTELY NOT. The number has to be EXACTLY in the middle of the high and low, that's how the high and low are determined. Therefore, the article's assertion that 98,000 was the middle, 8,000 was the low, and 194,000 the high GUARANTEES that at least one of the numbers in the article is incorrect. More importantly, all this suggests that the article was purposefully deceitful, trying to discredit the Lancet Study for using standard statistical methods, banking that most readers would be too unfamiliar with those methods to see through their deception. |
| ____________________________________ [May 18,2006 1:00am - PatMeebles ""] 1) Saying something is definitely in between 8,000 and 194,000, or in between 8% and 94%, is total unreliable garbage in terms of finding the final results, regardless of whether it's raw numbers or percentages. Regardless of which number came first, the fact is the certainty of it within a range is completely unreliable, too. 2) The reason why the numbers are in such a high range is when the people did the study, they surveyed people from certain areas, then assumed that the rest of the country was exactly the same. In other words, they assumed that the entire Kurdish North was exactly the same in terms of violence as Falluja. That's a ludicrous way of doing studies. Should I assume that the entirety of Massachusetts is as violent as Roxbury? Westborough is as violent as Springfield? Ridiculous. 3) Of course there would be a higher range if you want the certainty to be greater. But you have to have some point where you say "ok, maybe this range is too great." If their certainty was in a range of 30,000, there wouldn't be nearly as much criticism. |