Comments on: All models are wrong, but some are useful

By: brianISU

brianISU — Tue, 08 Jul 2008 13:32:13 +0000

I find it interesting that the guy is mentioning that models are becoming obsolete, yet he keeps mentioning the use of statistical method. How are statistical methods model free? Even nonparametric must make some assumptions. It seems to me that this essay is more about how models aren’t becoming obsolete, but are evolving to a new stage.

By: hlee

hlee — Wed, 02 Jul 2008 00:27:41 +0000

nice quote and I agree but stamp collecting is a very elaborative work once your collection goes beyond one binding. If stamp collecting is confined to a chronological assembly, I have nothing to say. Conversely, consider association rules for decision making. We need some relationships to lay out products in a store by looking at purchase patterns. I think correlation here has a more broad sense than astronomers’ perception (laying a straight line on a scatter plot with error bars to draw a coherent relationship). The quote belongs to the past, not the future that asks clusters of computers to handle data.

By: vlk

vlk — Tue, 01 Jul 2008 18:19:13 +0000

Allow me to disagree vehemently with the premise of that article. Correlation is never enough. Without theory and modeling, all you get is stamp collecting.