How To Misuse Statistics

Attila Dimedici writes "This story does a great job of showing how statistics are misused by comparing the incidence of voting for Democrats to the incidence of cancer. While the story has a strong ideological bias, it does a good job of poking fun at the way politicians (and others) misuse statistics."

Comments

Strong Ideological Bias?

[GiveBenADollar]

Ahh, but if a similar study was published against republicans it would be unbiased and informative about misuse of statistics. Here's an easy test for political bias. If a news agency lists a politician involved in a scandal and does not report (or buried in the story.) the political affiliation of the politician they are of party X. If the news agency blankets the story with party connections including ties to prominent political figures they are of party Y. Read/Watch for a while and you will quickly be able to fill in the X and the Y for most media outlets.

Re:Strong Ideological Bias?

[Daniel Dvorkin]

Fortunately, FOX News is always on the alert for suchg shenanigans, and takes prompt corrective action by clearly identifying nefarious Democrats such as Foley and Sandford.

Oh, wait a minute ...

Re:With all due respect and all that

[Daniel Dvorkin]

No, you made your point accurately. This isn't a story about the misuse of statistics; it's a story about the misuse of the concept "the misuse of statistics." TFA repeatedly claims that an absurd comparison of voting patterns and cancer rates is "just as valid" as studies showing a meaningful causal link between lack of insurance and death from lack of medical care -- but making that claim (and repeating it ad nauseam) doesn't make it true. The moral is not "stupid Democrats don't understand statistics," it's "stupid Republicans don't understand statistics and make fools of themselves when they try to make fun of Democrats who do."

There are three types of lies.

[dov_0]

Lies, damn lies and bloody statistics.

0% of people surveyed were surprised in any way by TFA.

Re:There are three types of lies.

[JWSmythe]

No, no, 100% of the sample group were surprised by TFA.

    Of course, "TFA" is the street name of our carefully selected poll taker. He's 6'5", a serious body builder, and covered in tattoos. He waited in a dark alley, and would jump out and ask the simple question "Are you surprised motherf***er?"

    Another thing we noted in the study was that conducting such a study can be very profitable. 60% of the people dropped their wallet or purse. The other 40% groveled before dropping their belongings.

    Next time, we're aiming for a larger sample group.

Re:There are three types of lies.

[IBBoard]

I don't believe you. Don't you know that 95% of statistics are made up on the spot?

Correlation implies correlation

[anglophobe_0]

I'm NOT trying to make it sound like I know what I'm talking about, because the following is entirely conjecture, with barely any thought behind it: I think everyone takes it for granted that urban areas are more likely to be blue than red (politically). Do various environmental factors also lead them to have higher cancer rates? I've not verified that, but I did just think it would be interesting to discover that the two maps aren't mere coincidence.

Re:Correlation implies correlation

[Hognoxious]

Well for one, too much sun exposure can cause skin cancer.

Sometimes you get a correlation but there's no apparent causal link. It can be that some third factor is influencing both the things you're measuring. You may have hit it with the urban/rural thing here.

I'd say though that there is a direct and somewhat obvious causal link between not having medical insurance and death rates...

Re:Correlation implies correlation

[anglophobe_0]

Check out this article on politifact: here. It doesn't dispute that people's quality of life suffers due to lack of healthcare, but it does dispute some misinformation regarding actual earlier deaths.

Personally, I'm not against poor people like myself having healthcare. If they can't get it through their employer, I would rather see private charity step in than the government, but unfortunately we do live in a country where the government has largely taken the place of private charity. What must be, must be.

Re:Correlation implies correlation

[Hognoxious]

Would you rather rely on private charity? It can be withdrawn on a whim. I bet it's looking pretty thin at the moment.

Now logic implies that if you don't get treatment for a fatal condition you'll die, and if you don't have the insurance to pay for the treatment you won't get it. So how can your assertion be true? Key phrase here:

the safety net of public hospitals and community clinics providing "'good enough' access to care for the uninsured to keep their mortality rate similar to that of the insured."

You're more or less doing it, but in an uneven, ad-hoc, and inefficient way.

One other thing. Why do you dismiss quality of life so lightly? If someone survives and makes it back to full health he can be a productive member of society again. Isn't that better for everyone than if he survives as an invalid?

Re:Correlation implies correlation

[anglophobe_0]

I don't mean to shrug off quality of life. I think it's very important. I do not believe there is any constitutional protection for someone's quality of life: "life, liberty, and the _pursuit_ of happiness". However, I do believe that we have a moral (not civic) duty to help those in need. So, I'm against public health care on the same grounds that I'm against government ownership of GM - that stuff isn't the government's job.

Actually, charitable giving has gone down due to the economic crisis, but probably not as much as you think: link. I believe it would go up if the government would tax us less...but then again maybe I'm being to optimistic to think people would act humanely without government coercion.

You talk like health care is a never-ending resource. I saw John Q, and I felt terrible, but isn't it true that there are more people in need of a heart transplant than there are hearts available? I would much rather have a private charity or even an HMO and hospital board deciding whether I deserve to be on the list, than a government-run board. We don't have to use vitriolic terms like "death panel", but the truth is that health care is "scarce" in the economic sense. Check out this article: link.

I have no problem with giving health care to people who need it, or with helping people who can't afford it. I just think the government will do just as bad a job with this as they have with so many other services, and the people who need to be helped won't really be helped in the long run. You say private charity is unreliable, but I saw too many people turn to the Salvation Army when OSHA and the Red Cross (I know, RC is not exactly government, but they're much more tied in) were turning people away after Katrina (I was living in Missouri at the time) to agree with you.

Re:Correlation implies correlation

[Hognoxious]

then again maybe I'm being to optimistic to think people would act humanely without government coercion.

I'd say more than a bit...

I just think the government will do just as bad a job with this as they have with so many other services

You don't think that's getting to be a bit of a tired old stand up comedian's cliché? Other countries seem to be able to do it. France's is of comparable quality but half the cost. The UK's NHS is lower quality, but considerably lower cost. And the latter was set up in a country that was still in ruins after a war.

You say private charity is unreliable, but I saw too many people turn to the Salvation Army when OSHA and the Red Cross (I know, RC is not exactly government, but they're much more tied in) were turning people away

That's not what I was talking about. I was referring to the fact that the funding could dry up at any time. I'm sure the Sally Army do their best, I have a lot of respect for them but they can't do miracles.

Re:Correlation implies correlation

[anglophobe_0]

Again, I think the fact that charitable giving has gone down so little over the past two years, compared with the downturn in tax revenue makes charitable giving MORE reliable than tax-funded charity.

My main justification for trusting people to give is that they already do. Would they give as much as the government currently takes? Probably not, to be fair. But I do believe the money would be used more efficiently, hopefully making up.

I know it's cliche to bash government inefficiency...but that's only because government efficiency has been so bad for so long. Private business has always been more efficient than government, which is why I'm for a freer market AND more private charity.

From a pragmatic standpoint, I think that there are valid arguments (although, I believe, not enough) for government-provided healthcare. From a philosophical standpoint, however, I see zero justification for expanding the government in any way. Anyone remember the parable of the snake and the mongoose?

As valid as anything else

[bluefoxlucid]

This is blatantly stupid; however, there's plenty of stuff that makes obvious and logical sense that's also wrong and is shown by incorrectly handled statistics to be correct. Like the Yoplait study about how yogurt makes you drop lard off your fat ass... a study on 10 people, unblinded, who probably were engaging in a healthier life style as well. No controls.

Re:As valid as anything else

[JWSmythe]

    I won't argue against what you're saying, but do you have a citation for that?

    There are an awful lot of urban legends going around, and I don't like getting caught repeating any of them as fact.

    Most statistics regarding weight loss usually carry the disclaimer "with proper diet and exercise." You could lose weight eating one candybar every day, if you backed that up with an otherwise good diet and exercise. :)

 

Re:As valid as anything else

[bluefoxlucid]

Yes that's the point. Statistics can be bent in obviously flawed ways; but we're able to make completely reasonable assertions and flash numbers to back them up pretty easy too, by misapplication. I mean hell, on one data set and one single analysis, you can show that there's more black drug dealers AND black drug dealers sell more drugs per drug dealer; but at the same time, white drug dealers overall sell more drugs. Then you can use the same study to argue that {blacks,whites} cause more drug crime. I've done exactly that before, just for fun.

Re:With all due respect and all that

[JWSmythe]

    But in some cases, the drugs only exist because of the condition, which was caused by modern lifestyles.

    I found this study which shows a marked increase in mortality due to diabetes between 1840 to 1970. Most of the increases happened between 1880 to 1911, when refined sugars and fatty foods were introduced. I've seen other studies that reflect this also.

    The cure could become virtually unnecessary if the cause was removed, rather than profiting from the treatment.