Otherwise, I think that the point of this post breaks down when one realizes that you are talking about two different types of models. One type is created as an ideal, perhaps we could term this a “prescriptive model.” This is the type that is airbrushed, idealized, etc. However, descriptive models are used to approximate the real world (with admittedly limited information, yet the best available information). The goals are different for these different models, and you are critiquing the latter as if it were the former.

You are absolutely correct that models are imperfect approximations of the world. That is precisely why we call them that: MODELS. I never claimed that my third grade diorama, or one might say, model, of the solar system approached some real replica of the actual solar system. However, given my skill and knowledge as well as the information available to me as a third grade student, my clay and paper replica was perhaps the best model allowed under my expertise, available materials, and knowledge of the subject. It was nonetheless imperfect.

In evaluating any model of any system, one must always take into account first and foremost the assumptions of the model, for there are always essential assumptions on which the predictions will vitally hinge (a point well summarized in the second source on our post, Leland Teschler’s editorial). One must also consider the data on which the model is based, and the quality of said data, as well as the actual structure of the model. The Watts data is certainly relevant in this vein, but people are also drawing very preliminary conclusions. It does not follow from his data that the U.S. temperature data is unreliable. This would have to be demonstrated by showing that now-compromised sensors have deviated from non-compromised sensors, which could be done fairly simply with (uh oh, here it goes) a regression model comparing the temperature data against Watts’s ratings. One other consideration is that human modification of the earth’s surface itself actually changes the reflective value of the surface. Would these encroachments that Watts talks about then provide some sort of approximation of this effect (albeit an overestimation?). I bring up these counterpoints because I think it is too quick (and perhaps politically motivated) to say that it logically follows that surface data should be thrown out. What about trends using only the sensors that have not been compromised?

There are certainly limitations to weather models as there are to economic models, but those that are constructed without obvious biases also give us the best possible predictions of the system, given available knowledge. To say that modeling itself is an all around fruitless endeavor is over-reaching, and perhaps ends the discipline of economics, as well as many others. I would hate to tell Brent that.

-Jacob

Harrison’s last blog post..Not Simply Nocturnal Emissions