Comments on: Lies, Doggone Lies, and Statistics

By: aaron

aaron — Wed, 09 Dec 2009 01:29:31 +0000

36%? seriously? it seems to me that if a relates to b less than half the time, that’s a pretty good sign that a and b aren’t really related. i would think it would need to be at least 75% to declare any sort of correspondence, and even that would be a bit shaky.

By: shack

shack — Thu, 03 Dec 2009 23:50:56 +0000

Great rant! Love it!

I am especially excited/challenged/frustrated with the last two sentences. It is difficult to hear and discern those voices who are unafraid to rigorously utilize the tool of science in trying to understand the universe who don’t make it a crusade against some “humanistic” theory.

Anyway, thanks for the article.

By: Jamie the Very Worst Missionary

Jamie the Very Worst Missionary — Tue, 01 Dec 2009 16:17:34 +0000

I completely agree with you, Christina! This is an issue of applied critical thinking. (Which you, Stephen, have done really well here.) Thanks!

By: Stephen Simpson

Stephen Simpson — Mon, 30 Nov 2009 21:42:40 +0000

Rob,

There is a very strong correlation between the way you wear your socks and my inability to make it through life without huffing glue.

Luv,
Steve

By: Jim

Jim — Mon, 30 Nov 2009 19:42:37 +0000

A great journalist/scientist calling out these faults is Ben Goldacre for The Guardian (a British news paper).
I concur, the problem is that loose connections are missunderstood by the news media who proclaim “truth” where even the original scientist wouldn’t, which then gets claimed by talk-show hosts to show how smart they are. But until we (the general public) call news media to account for this, nothing is going to change.
Statistics can be used to prove anything you want. 76% of people know that.
And anything can be claimed as truth, even bold-faced lies – just look at the Obama birthcitificate debacle.

By: Larry Shallenberger

Larry Shallenberger — Mon, 30 Nov 2009 19:22:21 +0000

Thanks for your example. It’s pretty standard “best practice” for pastors to ask applicants who want to work with children if they had been the victim of sexual abuse as a child. Insurance companies suggest this question to churches and churches comply, wanting to have good insurance rates and wanting to protect children.

I can’t ask that question though. Applying to work with children shouldn’t strip you of your dignity. We do reference checks, national sexually offenders database searches, county criminal records searches, applications, and interviews.

But I worked with enough abused kids as a mental health caseworker to not have the heart to subject them to a frivolous and hurtful line of questioning like this.

By: Strong Odors

Strong Odors — Mon, 30 Nov 2009 18:22:36 +0000

I clicked on this article solely on the splash image from the View.
i thought it was a funny picture.

glad i clicked it.

By: Christina

Christina — Mon, 30 Nov 2009 17:36:23 +0000

Thanks, Stephen, for a good rant.

I’d like to add that while this post is a “rant against bad science,” it is also a rant against poor critical thinking skills.

We should work toward a better understanding and application of scientific method, sure. But let’s apply the kind of thinking that prompts good questions to all areas of life.

By: Rob

Rob — Mon, 30 Nov 2009 17:22:48 +0000

Well said, Stephen. You can rant to me anytime.

Unless its about the way I wear my socks.

By: Nathan Bubna

Nathan Bubna — Mon, 30 Nov 2009 16:29:26 +0000

Too many people out there think science is somehow exempt from basic philosophy and thus disregard the lessons of Ecclesiates, Proverbs and Socrates. The only sane, fundamental assumption to any logical and/or scientific exercise is that we are ignorant and biased. Here are some helpful rules when approaching science:

1) A scientist is always biased in choosing what to study and how to study it. It is not possible to be unbiased, anyone who claims otherwise is a fool.
2) Everyone observing results of a study is biased in interpreting it, including (and often especially) the expert scientist in charge.
3) No amount of correlation EVER equals causation. Period. No exceptions. So long as you only have correlative evidence, your theory is unproven and alternate explanations are possible. Seriously, stop trying to think of exceptions because there are NONE.
4) Causation is really, really damn hard to prove. So it’s ok to use strong correlations with a plausible theory of causation sometimes. Such theories can be useful models even if we’re not actually sure they’re correct. Just don’t ever forget that the theory could be totally wrong regardless of how useful it is.
5) Having an experiment be “repeatable” is no substitute for actually having it be repeated.
6) Statistics are not data that results from a study. Statistics are always a biased interpretation of data that results from biased study (see rules 1 and 2). Further, statistics are usually among the most biased interpretations that can be derived from the data resulting from a study as they usually are far too succinct to be reliable.

That said, the scientific method is one of the most useful things we have out there for technological and medical advancement. It’s like democracy, a terrible form of government, but still the best one available to us.

This is why it is not enough to do a single, simple study and declare you’ve found truth.