June 16, 2006
WB: Math Error
Does the hack who wrote this story even know what the term "margin of error" means? Or does he think it's just the blank space at the top of all his term papers where his teachers used to write the "Fs"?
Posted by b on June 16, 2006 at 01:14 AM | Permalink
How the hell desperate for good news is the White House friendly press corps when they need to resort to spin as weak as this? Billmon didn't need to bold the offense; that one isn't even going to fool a freeper. I mean, honestly, they could have at least said "standard deviation" instead of "margin of error" if they were really trying to sell this to their slack jawed subscribers.
Posted by: Monolycus | Jun 16, 2006 2:09:42 AM | 1
Huh? The difference is maybe not statistically significant, but it means that it is more likely than not (66%, AIUI) that the rating has improved.
Wikipedia puts it simply and well:
When the purpose of polls is to compare percentages, the use of the margin of error is tempting, but is inappropriate if the two intervals overlap.
Posted by: jobo | Jun 16, 2006 3:53:49 AM | 2
I believe you're right, jobo. Differences in observed numbers between two polls are likely to indicate differences in the population polled, all else equal, even if the differences fall within the poll's margin of error.
From the same Wikipedia article you cite:
A web search of news articles using the terms "statistical tie" or "statistical dead heat" returns many articles that use these terms to describe reported percentages that differ by less than a margin of error. These terms are misleading; if one observed percentage is greater than another, the true percentages in the entire population are more likely ordered in the same way than not.
The attempt at spin is indeed desperate, but statistically I think Billmon is off base.
Posted by: maryp | Jun 16, 2006 7:42:26 AM | 3
Billmon is right.
You are trying to read noise.
Posted by: Gaianne | Jun 16, 2006 8:24:07 AM | 4
From a referral from the wikipedia article cited above :
Any generalizations obtained from a nonprobability sample must be filtered through one's knowledge of the topic being studied. Performing nonprobability sampling is considerably less expensive than doing probability sampling, but the results are of limited value.
I think this is much more relevant than the "margin of error". It seems much more likely to me that the stenographers at the MSM and their bosses who are looking for the "right" answer, whatever they may deem that to be in a given instance, are more likely to choose samples that are what the wikipedia authors term "convenience polls, judgemental sampling, or ad hoc quotas" than probability samples.
Posted by: John Francis Lee | Jun 16, 2006 8:26:05 AM | 5
Mathematics often gives me a real headache.
Posted by: Lewis Carroll | Jun 16, 2006 8:55:45 AM | 6
Here is a statistic that may be of value to study: (figures *do* have a "margin of error" of course):
There are over two hundred and ten million people in the United States over the age of eighteen and probably 90% or more are eligible to vote.Only between 25-30% vote Republican either for President or for Congress. In spite of the fact that *all* of the corporate MSM almost always show them in a favorable light while almost always showing the *others* in their skivvies (so to speak). How can that be overcome? How can all the *others* get enough of a toe-hold in Congress to change the election rules so that can't happen? Even if the rules are amended so that campaign funds are taxpayer funded, how do you prevent the president from spending millions and millions of dollars flying all over the country supposedly visiting military bases while *conveniently* campaigning nearby. How do you stop the president from using tax payer funds to travel all over the *world* on photo-ops on Air Force One and attendant security. The presidency has become a scripted stage show almost always *on tour* and constantly followed by the MSM paparazzi. As long as the main-(stream) *owners* in this so called ownership society control this situation it won't matter what the *polls* say. These no good wreckers and destroyers of anything decent will win.
Posted by: pb | Jun 16, 2006 12:48:13 PM | 7
Can a point or two difference in a poll really matter whatever the margin of error is?
Margin of error or no examining two points so close together and then extrapolating a direction from that can never work.
Yet the media do that all the time. One of my all time favourites was a local story on rates of food-poising from restaurants.
The health department had made significant changes to their regulatory regime and the number of people who notified a case of stomach upset, shits etc after eating at restaurants had dropped from about 800 a month to 250 a month over the space of the 18 months since the inspection regime had been altered.
Then along comes Joe journalists eager to do a story on danger, panic and incompetence. A sure fire seller.
So he bails up the scientist who designed and implemented the changes and says something like "Dr Blahblah since your policy has been implemented in the department of health, we have had several calls from people complaining they have gotten sick. How do you explain this?"
Dr Blahblah grins widely and pulls out his statistics and says "Look here up until 2003 notifications of food poisoning had averaged out at just over 800 a month for several years. Yet since we made the changes we only get around 250 cases a month. Here is the curve here."
The camera showed a wavering horizontal line that suddenly begins a swing downwards where it then settled into wavering along horizontally at a considerably lower level.
Joe journalist looks nonplussed for a moment then peers in at the bottom part of the graph, suddenly perks up and says "Decreasing? How do you explain this then Dr Blahblah? There you go misleading the public again. In April this year there were 232 notified cases of foodpoisoning from people who claimed to have eaten out, yet last month in May there were 247! You say they are decreasing when they are increasing. This is so typical of local government bureaucrats. I believe the public has a right to demand your resignation."
Resistance from Dr Blahblah was completely futile. I have no idea whether Joe journo was an asshole or just plain thick, but it didn't matter either way what the scientist said. As far as the journo was concerned it was game over. Food poisoning had risen therefore this alleged scientist was a liar and an incompetent.
It was easy to laugh at this until one heard so many people concerned about the new food poisoning epidemic.
Posted by: | Jun 18, 2006 8:05:20 AM | 8
I just heard an ad for a New Oreleans restaurant (Copeland's) praising its prime ribs: only 2 percent of a cow is prime, therefore prime rib is better than 98% of other cuts!
Posted by: Brian Boru | Jun 20, 2006 1:04:12 AM | 9