Abstract:
It is commonly-accepted wisdom that more information is better, and that information should never be ignored. Here we argue, using both a Bayesian and a non-Bayesian analysis, that in some situations you are better off ignoring information if your uncertainty is represented by a set of probability measures. These include situations in which the information is relevant for the prediction task at hand. In the non-Bayesian analysis, we show how ignoring information avoids dilation, the phenomenon that additional pieces of information sometimes lead to an increase in uncertainty. In the Bayesian analysis, we show that for small sample sizes and certain prediction tasks, the Bayesian posterior based on a noninformative prior yields worse predictions than simply ignoring the given information.
Citations
|
4930
|
Elements of Information Theory
– Cover, Thomas
- 1991
|
|
602
|
Bayesian Theory
– Bernardo, Smith
- 1994
|
|
193
|
The minimum description length principle in coding and modeling
– Barron, Rissanen, et al.
- 1998
|
|
148
|
Maxmin expected utility with a non-unique prior
– Gilboa, Schmeidler
- 1989
|
|
125
|
Applied Statistical Decision Theory
– Raiffa, Schlaiffer
- 1967
|
|
65
|
An invariant form for the prior probability in estimation problems
– Jeffreys
- 1946
|
|
63
|
Knowledge, probability, and adversaries
– Halpern, Tuttle
- 1993
|
|
47
|
On the interpretation of decision problems with imperfect recall
– Piccione, Rubinstein
- 1997
|
|
32
|
On Predictive Distributions and Bayesian Networks
– Kontkanen, Myllymki, et al.
|
|
31
|
Merging of opinions with increasing information
– Blackwell, Dubins
- 1962
|
|
19
|
Graphoid properties of epistemic irrelevance and independence
– Cozman, Walley
- 2001
|
|
19
|
On ambiguities in the interpretation of game trees
– Halpern
- 1997
|
|
14
|
Probability update: conditioning vs. cross-entropy
– Grove, Halpern
- 1997
|
|
10
|
Maximum entropy and the glasses you are looking through
– Grunwald
- 2000
|
|
9
|
On the suboptimality of the generalized Bayes rule and robust Bayesian procedures from the decision theoretic point of view
– Augustin
- 2003
|
|
9
|
On the principle of total evidence
– Good
- 1967
|
|
9
|
A contrast between two decision rules for use with (convex) sets of probabilities: γmaximin versus E-admissibility. Synthese
– Seidenfeld
- 2004
|
|
5
|
Divisive conditioning: Further results on dilation
– Herron, Seidenfeld, et al.
- 1997
|
|
3
|
Imprecise Probabilities and Their Applications
– Symp
|
|
3
|
A review of consistency and convergence rates of posterior distribution
– Ghosal
- 1998
|
