« Know your data 8: revealing the little choices in your life | Main | Prostate cancer screening under scrutiny »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Thomas Colthurst

I'm not sure why you say that Cowen and Sumner don't include any alternative models. Sumner states that level targeting "should be the standard model" in his post, and Cowen explicitly lists four alternatives here: http://marginalrevolution.com/marginalrevolution/2011/10/is-lm-keynesianism-why-not-and-which-alternatives.html

I also think that requiring that the alternative model be *provably better* than the critiqued model is an impossibly high standard outside of pure math. I think it is enough that the alternative be simply *more useful*. Sometimes this means preserving the useful parts of what is being criticized, sometimes it doesn't. The canonical example of the later is Copernican astronomy initially giving *less* accurate predictions than the Ptolemaic model.

Finally, if there is anyone who does not suffer from the belief that models can be proved wrong and discarded, it is Tyler Cowen. The man is methodologically pluralistic to a fault; his prose often reads like someone translating Bayesian model averaging into English. His critiques of IS-LM are best understood (well, by statisticians, at least :) ) as demonstrating a low value of P(model | evidence), particularly in the context of understanding contemporary problems.


Thomas: thanks for the comment. I did read the post on alternatives which came after my post. However, I don't see anything there. As some commenters on Tyler's blog point out, it's a hotchpotch of ideas that do not add up to a model. At least that is what it sounds like to me as an outsider (not an economist).
How do you show the alternative is "more useful" when you can't show it's "provably better"? I'm not sure I understand the difference.
The trouble I have is that whatever alternative Tyler is suggesting, a different researcher can go and list the assumptions that don't make sense, the features that are omitted, the predictions that fail, etc. That's because every model suffers from these types of issues.


Suppose a model, if applied, will kill a kitten you're fond of.
Suppose you can raise valid and persuasive criticisms this model but can't construct an alternative to it.
Why exactly is it the case that said criticisms can't be take seriously?

A pretense of knowledge can do real harm. It's better to be ignorant than to have faulty knowledge.

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep