The differential biology reader

 
« Back to blog

Be strategically wrong

Wrong Tomorrow attempts to bring accountability to pundits by tracking and verifying predictions. But this might not be possible in the general case if the experts are working against you:

Forecasting plays a vital role in human activity. Consumers, managers, and politicians make their decisions in part based on their anticipation of future events…The complexities in properly anticipating future events may encourage decision makers to seek experts' advice. The main difficulty is, however, that professional forecasts may not be reliable. If an expert is informed (i.e., he knows the relevant odds), then he can reveal the relevant probabilities to decision makers. However, if an expert is uninformed (i.e., he knows nothing about the relevant odds), then he may mislead the decision makers. A fundamental question is therefore how to determine whether experts are informed.
A test either rejects or does not reject each expert based on the observed data and the profile of the probabilities announced by the experts…However, consider the case in which all experts are uninformed (i.e., they do not know anything about true probabilities). We show that they can still independently produce false forecasts that are likely to both pass the test, no matter how the data evolve in the future. Hence, the data may not suffice to effectively discredit uninformed, but strategic, experts.

Olszewski and Sandroni. Manipulability of comparative tests. Published online before print. doi:10.1073/pnas.0812602106 PNAS 106 (13) 5029-5034

photo cc-by dsevilla

Comments (2)

Jun 04, 2009
nikete said...
The name of first author is Olszewski, without the a at the end. The seccond author is Sandroni without the b at the end.

A related ineresting result by them is

http://arxiv.org/abs/0904.0338

A nonmanipulable test
Authors: Wojciech Olszewski, Alvaro Sandroni
(Submitted on 2 Apr 2009)

Abstract: A test is said to control for type I error if it is unlikely to reject the data-generating process. However, if it is possible to produce stochastic processes at random such that, for all possible future realizations of the data, the selected process is unlikely to be rejected, then the test is said to be manipulable. So, a manipulable test has essentially no capacity to reject a strategic expert. Many tests proposed in the existing literature, including calibration tests, control for type I error but are manipulable. We construct a test that controls for type I error and is nonmanipulable.

Jun 04, 2009
"Olszewskia and Sandronib. Manipulability of comparative tests. Published online before print. doi:10.1073/pnas.0812602106 PNAS 106 (13) 5029-5034"

Thanks for the corrections on the authors' names. This was a careless copy–paste of the citation that took the author affiliation superscripts as well.

Leave a comment...