Has anyone read "200% of nothing"?

[Brainster]

I have found that it is extremely difficult to look at statistics and come to accurate conclusions. One trap is trying to get statistics to produce a certain result. Any fool can do that. Result - people on both sides of many debates say the statistics support their case.

There's an old saying that if you torture the statistics enough, they will confess to anything. Perhaps the most recent example in this regard is Bobby Kennedy Jr's piece in Rolling Stone, where he claimed that election officials in several small, rural counties switched 80,000 votes from Kerry to Bush, resulting in Bush winning Ohio and the election.

How did Kennedy arrive at this conclusion? By analyzing the result of the election for Ohio Chief Justice. Kerry did much better than a liberal black judge in most of the state, but in 12 rural counties she did better. If Kerry had done as well compared to that judge in those rural counties, he would have gotten 80,000 more votes in those counties (and Bush 80,000 fewer).

But if you accept Kennedy's logic, what happens? Well, for starters, obviously, Kerry would have won Ohio and the presidency. But second, Bush's share of the vote in those 12 counties would have declined dramatically compared to 2000. He would have gotten about 55% of the vote in those counties as compared to 67% in 2000.

Well, you can probably guess where I'm going with that, right? Were there other Ohio counties where Bush did 12 percentage points worse in 2004 than 2000? Answer: No, his worst decline was about 5 percentage points in any county.

So even though Kennedy "proved" statistically that Kerry won, it's not hard to prove statistically that there's no chance that outcome happened.

Moral of the story: If you go looking for a desired result in statistics, you can probably find it.

 

[Skeptic Ginger]

My opinion is that few people, including "rationalists" and "critical thinkers", take the time to examine the research data and methodology, and that even fewer look for corroborating data to support the conclusions. Unfortunately, this leads to reliance on broadcast media to interpret the data, which they are accustomed to do -- completely wrong. I recall a study comparing low carbohydrate and low fat diets. The study design had <100 participants all >40 BMI who were randomized to one of the treatment groups. The conclusions were mixed based on the limited number of participants, the high drop-out rate, and the limited study period (best as I can recall, 6 months). But the news media (a national one at that) reported the study as conclusive evidence that both diets were effective at improving a multitude of risk factors, but further reported that the low-carb diet was superior to the low-fat diet in improving glycemic control, even though the P value was >0.05.

BTW, I recently heard a statistic (on Dr. Jim Brackman's program) that coronary heart disease is no longer the dominant cause of mortality in the US. I tried finding mortality statistics but have had no luck. Anyone?

Sadly, it is a majority of uniformed that populate this country. I expect very few countries can claim otherwise. It amazes me we have such incredible advances in science given the above assumption.

But I do not give up. I am on a mission to teach people how to be informed. Perhaps I will influence a few, perhaps many, but there is no reason not to try.

As to cause of death, CDC reports numbers from 2002 in their 2005 summary. Even with infectious disease case reports they lag a few years behind with their official annual reports despite weekly information available in the MMWR. Number of deaths in 2002 in the USA for leading causes of death are as follows:

Heart Disease: 696,947
Cancer: 557,271
Stroke: 162,672
Chronic lower respiratory diseases: 124,816
Accidents (unintentional injuries): 106,742
Diabetes: 73,249
Influenza/Pneumonia: 65,681
Alzheimer's disease: 58,866
Nephritis, nephrotic syndrome, and nephrosis: 40,974
Septicemia: 33,865

Source: Deaths: Leading Causes for 2002, table 1

Infectious disease as a leading cause of death has been steadily increasing. I have not heard there have been substantial decreases in heart disease fatalities though we expect the statin drugs to have an impact soon.

 

[Pete Tattum]

Yes, I have a copy. It's very eye-opening...Innumeracy is dangerous.

 

[sickstan]

As to cause of death, CDC reports numbers from 2002 in their 2005 summary. Even with infectious disease case reports they lag a few years behind with their official annual reports despite weekly information available in the MMWR. Number of deaths in 2002 in the USA for leading causes of death are as follows:

Yeah, those are the most recent numbers I've found as well. I should contact Brackman to see where he got his info.

 

[Dustin Kesselberg]

Dewdney wrote a book about how people use statistics to lie? Well, he should know. His name comes up a lot in 9/11 conspiracy circles. In order to "prove" that passengers on the hijacked flights could not have made phone calls, he took some cell phones up to 8,000 feet in his Cessna over London, Ontario, and extrapolated that there is virtually no chance of a cell phone working at 32,000 feet. Case closed. (Never mind that many of the calls were made with the planes' installed Airphones.) Dewdney's "research" is used constantly by the conspiracy believers to show that the cell phone calls were faked. He called his cell phone escapade "Project Achilles." No bias there, huh?

This is an Ad Hominem attack on Dewdney. I don't know anything about his research into 9/11 phone calls, I don't know much about Dewdney himself. I picked up the book because I liked the content not because I liked the Author.

People change in 15 years(How long ago 200% of nothing was written). Right now he seems to be a Caucasian Muslim who thinks 9/11 was a Hoax. But when the book was written who knows?


The fact that he might be a kook today is irrelevant to the book '200% of nothing'. The book itself is very reliable and accurate and explains how math is abused in modern society.

 

[Dustin Kesselberg]

It's not that Statistics themselves are bad. It's how they can be manipulated to fit the interests of those manipulating them. One example..


Pepsi says "In recient taste tests, More people prefered the taste of Pepsi over Coke!"

What does this mean? Not enough information is given to make a conclusion. You would probably conclude that they had a random population drink either coke or pepsi and then choose.

However the statement would be true if they simply choose numerous groups of non-random tasters and then picked the group that fit their claim and ignored the others.

 

[Skeptic Ginger]

Yeah, those are the most recent numbers I've found as well. I should contact Brackman to see where he got his info.

I couldn't find a Dr Jim Brackman on Google. What program is it? I was going to see if there was any reference on the program's web page.