TEST YOUR KNOWLEDGE OF THE INCOME DISTRIBUTION

Think you know it all? A good deal of decision-making research centers around people's abilities to make accurate estimates and inferences.

Those who like to test their knowledge might be interested in this fun video game / Web experiment put together by Decision Science News and Lionel Page.

In it, you get to enter your beliefs about the inequality of income in the USA, and at the end, you can find out how accurate you were. Fun!

Give it your best shot at: http://www2.decisionresearchlab.com/db/hi/

A TALE OF TWO SELVES

Psychologists and economists love to talk about the notion of two selves: present self and future self. It's a nice way to explain the tendency to have one preference about the future, but a very different preference when the future becomes the present. On Sunday, future self might want to go to bed early on Thursday, wake up early on Friday, and hit the gym where it will listen to one hour of "Listen-and-repeat Italian" lessons while mastering the StairMaster.

However, come Thursday evening's dinner with a client, this voice cannot be heard next to that of present self saying yes to dessert, coffee, after-dinner liquer, and a postprandial visit to the pub. Sunday's voice is also asleep Friday morning, when the present self resets the alarm from 5:30 to 8:00.

In a clever April fools joke, the website www.thinkgeek.com proposed a solution in the form of an alarm clock that donates money to your most-hated cause should you hit the snooze button. Imagine giving money to a despised politician every time you slept in. Might that get you out of bed?

While the SnuzNLuz alarm clock was a joke, it was a brilliant one. I believe that someone will run with this or a very similar idea. Many future selves find their present selves to be their own worst enemies, and might be willing to pay obedience school tuition. In fact, the much-buzzed about www.stickk.com is built on this model: taxing yourself for failing to lose weight, quit smoking, etc.

I am reminded of the time I was a postdoc at Columbia University, on the job market, and deep in a publish-or-perish the phase of my career. I instituted a similar (though lower-tech) mechanism. My rule was that if I didn't write a certain number of pages each day, I would lose five dollars. I think I lost about $60 on the scheme, though it did land me a job I love.

I remember being seriously conflicted about whom to give the money to if I procrastinated. I felt that if I gave it to a good cause, I would be continually justifying my procrastination as charitable. I felt that if I gave it to a bad cause, that would be evil. I also feared that I would start justifying my procrastination by telling myself the bad cause isn't so bad. (Sound far-fetched? The idea that we might infer our preferences from our actions is a key, if not field-defining, idea from social psychology.)

In the end, I chose to leave the money on a seat on the New York subway. Maybe a good person would find it, maybe a bad person would find it, all I was certain of was regretting my procrastination. Given that you're not evil, if you found $5 on the 1/9 train around 2005, I hope that it inched you closer to your goals.

MAJOR CHOICE STRATEGIES

My somewhat annotated copy of Hastie & Dawes

Reid Hastie and Robyn M. Dawes, in their classic Rational Choice in an Uncertain World (pp. 232-234), outline some "major choice strategies," stemming from several schools including the Heuristics and Biases, Adaptive Decision Maker, and Fast and Frugal research programs:

Strategy: DOMINANCE
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	LOW	NON-COMPENSATORY	ALTERNATIVE	YES
"Search for an alternative that is at least as good as every other alternative on all important attributes and choose it or find an alternative that is worse than any other alternative on all attributes and throw it out of the choice set."
Strategy: ADDITIVE LINEAR (Multi-Attribute Utility Theory)
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	V. HIGH	COMPENSATORY	ALTERNATIVE	YES
"Weight all the attributes by their importance (with reference to the current goals of the decision maker). Then consider each alternative one at a time and calculate a global utility by valuing each attribute, weighting it by its importance, and adding up the weighted values."
Strategy: ADDITIVE DIFFERENCE
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	V. HIGH	COMPENSATORY	ATTRIBUTE	YES
"Consider two alternatives at a time; compare attribute by attribute, estimating the difference between the two alternatives; and sum up the differences across the attributes to provide a single overall difference score across all attributes for that pair. Carry the winner of this comparison over to the next viable alternative and make the same comparison. At the end of this process, the best alternative is the one that has 'won' all the pairwise comparisons."
Strategy: SATISFICING (CONJUNCTIVE)
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	LOW	NON-COMPENSATORY	ALTERNATIVE	NO
"First set 'acceptability' cutoff points on all important attributes; then look for the first alternative that is at least as good as the cutoff values on all important attributes or use the strategy to select a set of good-enough alternatives (all above the cutoff points) for further consideration."
Strategy: DISJUNCTIVE
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	LOW	NON-COMPENSATORY	ALTERNATIVE	NO
"First, set 'acceptability' cutoff points on the important attributes; then look for the first alternative that is at least as good as the cutoff value on any attribute or use the strategy to select a set of alternatives that are very good on at least one dimension for further consideration."
Strategy: LEXICOGRAPHIC (AND TAKE-THE-BEST)
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	MEDIUM	NON-COMPENSATORY	ATTRIBUTE	NO
"First, review the attributes and pick the one most important attribute; then choose the beest alternative on that attribute. If there are several "winners" on the first attribute, go on to the next most important attribute and pick the best remaining alternative(s) on that attribute. Repeat until only one alternative is left ... [Similar to the] take-the-best fast-and-frugal heuristic (successful in choice and judgment environments that reflect the distributions of alternatives and attribute values in real, everyday environments). The only adjustment to our description would be to substitute the word 'validity' (predictive accuracy) for 'importance'; order the attributes considered by their past validity in discriminating between good and bad alternatives."
Strategy: ELIMINATION BY ASPECTS
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	MEDIUM	NON-COMPENSATORY	ATTRIBUTE	NO
"Pick the first attribute that is salient and set a cutoff 'acceptability' point on that attribute. Throw out all alternatives that are below the cutoff on that one attribute. Then pick the next most attention-getting attribute, set an 'acceptability' cutoff on that attribute, and again throw out all alternatives that are below the cutoff. Repeat until only one alternative is left."
Strategy: RECOGNITION HEURISTIC
	Mental Effort	Compensatory vs. Noncompensatory?	Whole vs. Part	Exhausive?
	LOW	NON-COMPENSATORY	ALTERNATIVE	NO
"In some choices, people are so poorly informed about the alternatives that they simply rely on 'name recognition.' They choose the first alternative that they recognize ... in many realistic choices and judgments the 'fast and frugal' recognition choice heuristic behaves surprisingly well."

Source: Hastie, Reid & Dawes, Robyn M. (2001). Rational choice in an uncertain world. Sage: Thousand Oaks, CA, pp. 232-234.

About the Authors:

REID HASTIE

Reid Hastie is a Professor of Behavioral Science on the faculty of the Graduate School of Business in the Center for Decision Research at the University of Chicago. His primary research interests are in the areas of judgment and decision making (legal, managerial, medical, engineering, and personal), memory and cognition, and social psychology. He is best known for his research on legal decision making (Social Psychology in Court [with Michael Saks]; Inside the Jury [with Steven Penrod and Nancy Pennington]; and Inside the Juror [edited]) and on social memory and judgment processes (Person Memory: The Cognitive Basis of Social Perception [several co-authors]). Currently he is studying: the role of explanations in category concept representations (including the effects on category classification, deductive, and inductive inferences); civil jury decision making; the role of frequency information in probability judgments; and the psychology of reading statistical graphs and maps.

Reid Hastie vita

ROBYN M. DAWES

Robyn Dawes is the Charles J. Queenan, Jr. University Professor Ph.D.: University of Michigan Department Member Since: 1985 at the Department of Social & Decision Sciences at Carnegie Mellon University. His research interests spans five areas: intuitive expertise, human cooperation, retrospective memory, methodology and United States AIDS policy. He states, "I write journal articles and books because I believe the information they contain could be valuable -- at least on a "perhaps, maybe" basis. I have never written anything with the expectation that it will sell, or become a "citation classic" (although one of my articles has). I believe that in American culture we are obsessed with outcomes rather than with behaving in ways that tend to bring about the best expected outcomes, while "time and chance" play a very important role. [...] Some of my clinical colleagues claim that feelings are not understood until they can be put into words. My own view is that every translation of a feeling, thought, idea or mathematical form into words involves at least a small element of automatic distortion, often a much larger element."

POLITIMETRICS

Lionel Page (University of Westminster), in conjunction with Paul Antoine Chevalier (Paris School of Economics), Dan Goldstein (London Business School & Decision Science News), Leighton Vaughan Williams (Nottingham Trent University), and Peter Urwin (University of Westminster) are pleased to bring you Politimetrics.com a Web site that uses prediction markets to forecast election outcomes and more.

What candidate would have the highest probability to win the presidential election if nominated?

What candidate has the program which is more likely to foster growth, reduce unemployment and crime?

All these crucial questions for the voters in this 2008 US election cannot be answered using traditional polls. Using prediction markets in a innovative way, the website Politimetrics.com proposes answers to these questions.

Politimetrics.com presents the estimation of the the conditional probability of success of each candidate if nominated/elected. The numbers are estimated in real time, directly from the prices on specific Intrade contracts.

At this stage of the primary campaign, politimetrics proposes the best answer available the question: "who are the candidates the most likely to win the presidential election if nominated?" Later on in the campaign, we plan to present an even more interesting answer: "which candidate would be the most successful president on a list of issues?" To do so, we have proposed to Intrade a series of specific contracts (listed under the section "Impact of Next President" on their website). Eventually, we hope to be able to answer to questions like:

Is Hilary Clinton more likely to be more effective in managing the economy than John McCain?

Is Mitt Romney more likely to decrease crime than Barack Obama?

Here a video tutorial on how to get started using the R Language for Statistical Computing.

http://www.decisionsciencenews.com/?p=261

The tutorial is best viewed in your browser’s full-screen mode, try pressing F11 in Windows.

R is free and open source, and constantly being improved upon by countless contributors worldwide. DSN highly recommends using R.

Topics covered include:

* Downloading and installing R in Windows * The R graphical user interface * Viewing the graphics demo * Vectors and basic stats * Simple plotting

http://www.decisionsciencenews.com

Leading Judgment and Decision-Making scholar Gerd Gigerenzer was interviewed in Salon.com this week about his new book Gut Feelings: The Intelligence of the Unconscious in a piece amusingly titled "Should National Security Depend on Michael Chertoff's Gut?"

The interview, which covers September 11th, fly balls, high school dropouts, illegal drugs, and the prostate, makes for attractive summer reading.

www.decisionsciencenews.com

ARGUMENTS AGAINST ALL SIGNIFICANCE TESTS

This week, the always-provocative J. Scott Armstrong submits this comment to Decision Science News:

"About two years ago, I was a reasonable person who argued that tests of statistical significance were useful in some limited situations. After completing research for "Significance tests harm progress in forecasting" in the International Journal of Forecasting, 23 (2007), 321-327, I have concluded that tests of statistical significance should never be used. Here is the abstract:

I briefly summarize prior research showing that tests of statistical significance are improperly used even in leading scholarly journals. Attempts to educate researchers to avoid pitfalls have had little success. Even when done properly, however, statistical significance tests are of no value. Other researchers have discussed reasons for these failures. I was unable to find empirical evidence to support the use of significance tests under any conditions. I then show that tests of statistical significance are harmful to the development of scientific knowledge because they distract the researcher from the use of proper methods. I illustrate the dangers of significance tests by examining a re-analysis of the M3-Competition. Although the authors of the re-analysis conducted a proper series of statistical tests, they suggested that the original M3-Competition was not justified in concluding that combined forecasts reduce errors, and that the selection of the best method is dependent on the selection of a proper error measure; however, I show that the original conclusions were correct. Authors should avoid tests of statistical significance; instead, they should report on effect sizes, confidence intervals, replications/extensions, and meta-analyses. Practitioners should ignore significance tests and journals should discourage them. http://dx.doi.org/10.1016/j.ijforecast.2007.03.004

The paper is followed by commentaries by Keith Ord, Herman Stekler, and Paul Goodwin, and by my reply "Statistical significance tests are unnecessary even when properly done and properly interpreted: Reply to commentaries" , which can be found online at http://dx.doi.org/10.1016/j.ijforecast.2007.01.010

This is happy news for practitioners, researchers, and students. On the other hand, it might create anguish among faculty who teach people about statistical significance."

http://www.decisionsciencenews.com
http://www.dangoldstein.com

IMPACTFUL, IMPROBABLE EVENTS CAPTURE THE IMAGINATION

Attention is turning towards the low-probability events that can change the world.

Institutions are undergoing epidemic preparations. Harvard Business Review has an feature on pandemics with two sections written by noted decision science researcher Baruch Fischhoff. A podcast is available here.

A book on high-impact black swans debuted at number five on the NY Times best seller list.

Popular television shows 24 depict again and again how a few individuals, without special titles or power, can topple governments and kill millions.

Decision researchers are investigating how people infer the likelihood of low-probability events from experience and description.

It is a pandemic pandemic?

Decision Science News

A new manuscript by Nassim and me

Comments welcome!

www.decisionsciencenews.com

A CHANGE OF WEIGHT REVEALS NEWTON'S FLAW

Interestingly, Newton's enumerated solution to Pepys' problem is correct (see previous DSN entry), but the logic is wrong, as Statistician Stephen Stigler points out.

The problem is now solved by with bionomial distribution: the probability of A is the greatest. For those who speak R, the probabilities are

Probability of Event A = sum(dbinom(1:6,6,1/6)) = 0.67
Probability of Event B = sum(dbinom(2:12,12,1/6)) = 0.62
Probability of Event C = sum(dbinom(3:18,18,1/6)) = 0.60

This was Newton's logic:

"If the question be thus stated, it appears by an easy computation that the expectation of A is greater than that of B or C; that is, the task of A is the easiest. And the reason is because A has all the chances of sixes on his dyes for his expectation, but B and C have not all the chances on theirs. For when B throws a single six or C but one or two sixes, they miss of their expectations."

But Stigler points out Newton's logic doesn't hold if we use loaded dice, in which the probability of a six is not 1/6 but 1/4.

Probability of Event A' = sum(dbinom(1:6,6,1/4)) = 0.82
Probability of Event B' = sum(dbinom(2:12,12,1/4)) = 0.84
Probability of Event C' = sum(dbinom(3:18,18,1/4))= 0.86

Mathemagically, the probabilities grow from A to C in this case.

For a fun and short read, check out:

Stigler, Stephen M. (2006). Isaac Newton as a Probabilist. Statistical Science, 21(3), 400-403. [Download]
Photo credit: http://www.flickr.com/photos/awshots/352212946/

www.decisionsciencenews.com