Wholesome 8===D

http://www.stat.columbia.edu/~cook/movabletype/archives/2009/07/work_with_me_in.html

Among other things, while on sabbatical in Paris next year I'll be working with my longtime collaborator Frederic Bois, a toxicologist who uses hierarchical Bayes models extensively. We have a project in toxicology that necessarily also involves research in Bayesian computation.

And, there's a postdoctoral position available! Here are the details:

Post-Doctoral Position in Systems Biology

Context:

The newly created Chair in Mathematical Modeling for Predictive Toxicology and Ecotoxicology at Compiègne University of Technology (UTC), and INERIS (French National Institute for Industrial Environment and Risks), collaborate on the development of Systems Biology tools and concepts for predictive toxicology. An important part of the on-going work focuses on the development of ab initio methods for modeling drug-drug interactions and metabolic interactions in complex mixtures of chemicals. Development and application of Bayesian numerical techniques to pathway identification and parameterization is another research theme. The research is performed in co-operation with the Biomechanics and Bioengineering laboratory (UMR6600) at UTC, which develops bio-artificial organs, to mimic in vitro the in vivo conditions. Research links also exist with a large consortium of laboratories working in the framework of the European Integrated Projects "PREDICT-IV" and "2-FUN".

Project:

The goal of the post-doctoral position now opened is to further develop, validate and illustrate an original approach for automatic integration of SBML and physiologically-based models for many-reactant metabolic / signaling / transcription /regulation schemes. Issues of parameter estimation and structural identifiability will be treated as needed. Under the supervision of the chairperson, the candidate will in particular:
- Participate in the development of the GNU MCSim software,
- Further the development of theoretical foundations of the approach developed,
- Develop case studies and applications,
- Propose the design of needed experiments and supervise the data collection,
- Actively publish in peer reviewers journals,
- Disseminate the results to the pharmacology, toxicology and systems biology research communities and industrial partners,
- Supervise students working on the project.

Candidate profile:

We are seeking a highly motivated individual interested in collaborating with an interdisciplinary group of modelers, biologists and bioengineers. A Ph.D. in computational biology, bioengineering, biomathematics or related fields is essential. Communication skills and fluency in English are a requirement. Applications include a statement of interest, CV, and names and contact for two academic references. Applications or requests for further information should be sent to Frederic Y. Bois (Frederic.Bois@ineris.fr).

Other information:

Salary: about 2500 € (gross) per month

Location: Compiègne University of Technology and INERIS, Verneuil-en-Halatte, Oise, France (60 kilometers North of Paris)

Duration: 24 months

http://www.stat.columbia.edu/~cook/movabletype/archives/2009/07/rubinism_separa.html

In the most recent round of our recent discussion, Judea Pearl wrote:

There is nothing in his theory of potential-outcome that forces one to "condition on all information" . . . Indiscriminate conditioning is a culturally-induced ritual that has survived, like the monarchy, only because it was erroneously supposed to do no harm.

I agree with the first part of Pearl's statement but not the second part (except to the extent that everything we do, from Bayesian data analysis to typing in English, is a "culturally induced ritual"). And I think I've spotted a key point of confusion.

To put it simply, Donald Rubin's approach to statistics has three parts:

1. The potential-outcomes model for causal inference: the so-called Neyman-Rubin model in which observed data are viewed as a sample from a hypothetical population that, in the simplest case of a binary treatment, includes y_i^1 and y_i^2 for each unit i).

2. Bayesian data analysis: the mode of statistical inference in which you set up a joint probability distribution for everything in your model, then condition on all observed information to get inferences, then evaluate the model by comparing predictive inferences to observed data and other information.

3. Questions of taste: the preference for models supplied from the outside rather than models inspired by data, a preference for models with relatively few parameters (for example, trends rather than splines), a general lack of interest in exploratory data analysis, a preference for writing models analytically rather than graphically, an interest in causal rather than descriptive estimands.

As that last list indicates, my own taste in statistical modeling differs in some ways from Rubin's. But what I want to focus on here is the distinction between item 1 (the potential outcomes notation) and item 2 (Bayesian data analysis).

The potential outcome notation and Bayesian data analysis are logically distinct concepts!

Items 1 and 2 above can occur together or separately. All four combinations (yes/yes, yes/no, no/yes, no/no) are possible:

- Rubin uses Bayesian inference to fit models in the potential outcome framework.

- Rosenbaum (and, in a different way, Greenland and Robins) use the potential outcome framework but estimate using non-Bayesian methods.

- Most of the time I use Bayesian methods but am not particularly thinking about causal questions.

- And, of course, there's lots of statistics and econometrics that's non-Bayesian and does not use potential outcomes.

Bayesian inference and conditioning

In Bayesian inference, you set up a model and then you condition on everything that's been observed. Pearl writes, "Indiscriminate conditioning is a culturally-induced ritual." Culturally-induced it may be, but it's just straight Bayes. I'm not saying that Pearl has to use Bayesian inference--lots of statisticians have done just fine without ever cracking open a prior distribution--but Bayes is certainly a well-recognized approach. As I think I wrote the other day, I use Bayesian inference not because I'm under the spell of a centuries-gone clergyman; I do it because I've seen it work, for me and for others.

Pearl's mistake here, I think, is to confuse "conditioning" with "including on the right-hand side of a regression equation." Conditioning depends on how the model is set up. For example, in their 1996 article, Angrist, Imbens, and Rubin showed how, under certain assumptions, conditioning on an intermediate outcome leads to an inference that is similar to an instrumental variables estimate. They don't suggest including an intermediate variable as a regression predictor or as a predictor in a propensity score matching routine, and they don't suggest including an instrument as a predictor in a propensity score model.

If a variable is "an intermediate outcome" or "an instrument," this is information that must be encoded in the model, perhaps using words or algebra (as in econometrics or in Rubin's notation) or perhaps using graphs (as in Pearl's notation). I agree with Steve Morgan in his comment that Rubin's notation and graphs can both be useful ways of formulating such models. To return to the discussion with Pearl: Rubin is using Bayesian inference and conditioning on all information, but "conditioning" is relative to a model and does not at all imply that all variables are put in as predictors in a regression.

Another example of Bayesian inference is the poststratification which I spoke of yesterday (see item 3 here). But, as I noted then, this really has nothing to do with causality; it's just manipulation of probability distributions in a useful way that allows us to include multiple sources of information.

P.S. We're lucky to be living now rather than 500 years ago, or we'd probably all be sitting around in a village arguing about obscure passages from the Bible.

http://feedproxy.google.com/~r/scienceblogs/YsBw/~3/jR1-CPQ6v5s/selective_aphasia_in_a_brain_damageed_bilungual_patient.php

http://scienceblogs.com/neurophilosophy/2009/07/selective_aphasia_in_a_brain_damageed_bilungual_patient.php

IN THE 1860s, the French physician Paul Broca treated two patients who had lost the ability to speak after suffering strokes. When they died, he examined their brains, and noticed that both had damage to the same region of the left frontal lobe. About a decade later, neuropsychiatrist Carl Wernicke described a stroke patient who was unable to understand written words or what was said to him, and later found in this patient's brain a lesion towards the back of the left temporal lobe. 

Thus the classical model of the neurological basis, of language being localized to two specific areas of the left hemisphere, was established. Recently though, researchers have found evidence that some components of language are encoded in other brain regions. Furthermore, it is still unclear how the brain represents language in bilingual people. Some studies suggest that both languages are represented in the same set of laguage areas, while others point to distinct neural substrates for the first and second languages.

A unique case study published in the open access journal Behavioral and Brain Functions sheds some light on this matter. The study, by Raphiq Ibrahim, a neurologist at the University of Haifa, describes a bilingual Arabic-Hebrew speaker who incurred brain damage following a viral infection. Consequently, the patient experienced severe deficits in Hebrew but not in Arabic. The findings support the view that specific components of a first and second language are represented by different substrates in the brain.

Read the rest of this post... | Read the comments on this post...

http://plato.stanford.edu/entries/perception-disjunctive/

[New Entry by Matthew Soteriou on July 10, 2009.]
Perceptual experiences are often divided into the following three broad categories: veridical perceptions, illusions, and hallucinations. For example, when one has a visual experience as of a red object, it may be that one is really seeing an object and its red colour (veridical perception), that one is seeing a green object (illusion), or that one is not seeing an object at all...

http://plato.stanford.edu/entries/cusanus/

[New Entry by Clyde Lee Miller on July 10, 2009.]
Arguably the most important German thinker of fifteenth century, Nicholas of Cusa (1401 - 1464) was also an ecclesiastical reformer, administrator and cardinal. His life-long effort, as canon law expert at Church councils, as legate to Constantinople and later to German dioceses and houses of religion, as bishop in his own diocese of Brixen, and as advisor in the papal curia, was to reform...

http://plato.stanford.edu/entries/plato-ethics-politics/

[Revised entry by Eric Brown on July 10, 2009.
Changes to: Main text, Bibliography]
Plato's Republic centers on a simple question: is it always better to be just than unjust? The puzzles in Book One prepare for this question, and Glaucon and Adeimantus make it explicit at the beginning of Book Two. To answer the question, Socrates takes a long way around, sketching an account of a good city on the grounds that a good city would be just and that defining justice as a virtue of a...

http://feedproxy.google.com/~r/The/Swamp/~3/2djbwlORfKk/

http://www.guzzlingcakes.com/?p=5742

I’ve been blogging for 10 years.

I couldn’t find any animated cake gifs that were hilarious enough, but I did find lots of marios. And so, to celebrate my Blogiversary, you are getting marios:

Ten marios each, for each of the 10 years I have been blogging. And in honour of 1999: Booyah!

http://feedproxy.google.com/~r/scienceblogs/wDAM/~3/ADbXPRiB4_I/swoopo.php

http://scienceblogs.com/cortex/2009/07/swoopo.php

Over at The Big Money, Mark Gimein has a fascinating article on Swoopo.com. Gimein calls Swoopo "the crack cocaine of auction sites" and says it's "the evil bastard child of game theory and behavioral economics." The site works like this:

Consider the MacBook Pro that Swoopo sold on Sunday for that $35.86. Swoopo lists its suggested retail price at $1,799; judging by the specs, you can actually get a similar one online from Apple (AAPL) for $1,349, but let's not quibble. Either way, it's a heck of a discount. But now look at what the bidding fee does. For each "bid" the price of the computer goes up by a penny and Swoopo collects 60 cents. To get up to $35.86, it takes, yes, an incredible 3,585 bids, for each of which Swoopo gets its fee. That means that before selling this computer, Swoopo took in $2,151 in bidding fees. Yikes.

In essence, what your 60-cent bidding fee gets you at Swoopo is a ticket to a lottery, with a chance to get a high-end item at a ridiculously low price. With each bid the auction gets extended for a few seconds to keep it going as long as someone in the world is willing to take just one more shot. This can go on for a very, very long time. The winner of the MacBook Pro auction bid more than 750 times, accumulating $469.80 in fees.

Gimein emphasizes the "sunk cost" illusion as motivating repeated bidding on Swoopo. Because we've already bid a hundred times on the laptop, we assume we're more likely to win on the 101st bid. But that's false; those earlier bids are utterly irrelevant. And then there's the jackpot effect: people tend to fixate on the big prize (the cheap laptop, the lottery payout, etc.) and not pay attention to the incremental cost of participation. Each bid, after all, is only sixty cents, but we could save hundreds of dollars on that Apple laptop! Unfortunately, the brain is terrible at addition, and so we quickly lose track of how much money we've actually spent trying to acquire the prize. (It also helps that Swoopo requires people to prepay for "bids" in packs, which makes it even easier for us to not think about bids in terms of real money. If Swoopo bidders had to use physical coins for every bid, I imagine the site would be far less appealing.)

But I still don't think that explains Swoopo's addictive appeal. Instead, I think the real appeal of the website is the sheer uncertainty. As an item nears the end of bidding, a big countdown clock appears. At any moment, someone else can come up in and bid on the item, which then resets the clock to twenty seconds. The process repeats and repeats, until the price gets to a point that discourages other bidders. (It's probably less discouraging to you, since you've already sunk $50 in bidding fees.) But here's the dirty secret of the site: after placing a bid, you're forced to wait and watch. You have no way of knowing if your bid will win, or if someone else will swoop in and bid on the laptop at the last possible second. In other words, it's just like a slot machine: you put in a quarter and wait for the wheels to whirr. With swoopo, the random number generator is other people.

Why is this excruciating state so appealing? The answer, I think, has a lot to do with how our brains process rewards. The popular myth of dopamine is that the neurotransmitter equals pleasure, that it's the hedonist chemical responsible for sex, drugs and rock n' roll. The dopaminergic reality, not surprisingly, is actually much more complicated. Consider the landmark work of Wolfram Schultz. His experiments followed a simple protocol: He played a loud tone, waited for a few seconds, and then squirted a few drops of apple juice into the mouth of a monkey. While the experiment was unfolding, Schultz was probing the dopamine-rich areas of the monkey brain with a needle that monitored the electrical activity inside individual cells. At first the dopamine neurons didn't fire until the juice was delivered; they were responding to the actual reward. However, once the animal learned that the tone preceded the arrival of juice -- this requires only a few trials -- the same neurons began firing at the sound of the tone instead of the sweet reward. And then eventually, if the tone kept on predicting the juice, the cells went silent. They stopped firing altogether. Schultz calls these cells "prediction neurons," since they are more concerned with predicting rewards than actually receiving them.

What's interesting about this system is that it's all about expectation. Our dopamine neurons constantly generate patterns based upon experience: if this, then that. They realize that the tone predicts the juice, or that betting on the laptop might get us a discounted reward. This means that our dopamine circuitry isn't just titillated when we win the auction - those predictive cells are excited every time we bid, as they wait to see whether or not the reward will arrive. Luke Clark, a neuroscientist at Cambridge, has done a very clever experiment that captures this process at work. Vaughan Bell, over at MindHacks, summarizes the results of the experiment:

The research team looked at the activity differences in the dopamine-rich mesolimbic system in a gambling task - comparing wins, misses and near-misses. Near-misses were where the reels on a slot machine just missed the payout.

It turns out that near-misses activate almost exactly the same dopamine circuits as actual wins - but here's the punchline - they were subjectively experienced as the most unpleasant outcome, even worse than total misses.

In other words, the dopamine system was firing like a rocket display but the experience was awful.

Interestingly, although near-misses were experienced as aversive they increased the desire to play the game but only when the person had some perception of control, by choosing what the 'lucky' picture would be. Of course, like choosing 'heads or tails', it's only an illusion of control because the outcome is random anyway.

This, in a nutshell, is how Swoopo works. It's one near-miss after another, as we bid and then bid again. The experience feels awful - we know we're wasting money - and yet we can't look away.

Read the comments on this post...

http://www.shockmd.com/2009/07/10/brain-rules-a-book-review/

http://www.shockmd.com/?p=3063

Brain Rules is the title of an excellent book. If your new to neuroscience of “brain science” this book is an excellent starting point. In a clear and funny way it explains complicated brain functions and their use in every day life. I read it with much pleasure. The author, John Medina uses entertaining examples and stories to substantiate the recent results of neuroscience research without loosing nuance.

How do we learn? What exactly do sleep and stress do to our brains? Why is multi-tasking a myth? Why is it so easy to forget—and so important to repeat new knowledge?

You can buy the book with a DVD.

The Brain Rules DVD is a lively tour of the 12 Brain Rules. You will experience firsthand Medina’s rare gift for making science fun, accessible, and relevant. Watch the DVD as an introduction to the book. Special features include: Short clips from all 12 Brain Rules, including the 3 p.m. nap zone, the Jennifer Aniston neuron, death by PowerPoint, and more Bonus material for business leaders and teachers, MP3s from the Brain Rules audio book, NTSC All Region

The book is accompanied by a website with a lot of information.

Related posts on this blog:

The Switch That Lifts Depression, from the Best of the Brain, book review. This book needs some preliminary knowledge.

http://www.stat.columbia.edu/~cook/movabletype/archives/2009/07/the_obama_admin.html

A websearch turned up this link to our report on Jeff and Justin's research. It's great to see this stuff out there, but, really, "LGBTQI"? The way things are going, we'll be going through the whole alphabet soon! There's gotta be another way. Once you have "Q" in there, doesn't that pretty much cover all the contingencies?

http://plato.stanford.edu/entries/friendship/

[Revised entry by Bennett Helm on July 9, 2009.
Changes to: Main text, Bibliography]
Friendship, as understood here, is a distinctively personal relationship that is grounded in a concern on the part of each friend for the welfare of the other, for the other's sake, and that involves some degree of intimacy. As such, friendship is undoubtedly central to our lives, in part because the special concern we have for our friends must have a place within a broader set of concerns, including moral...

http://plato.stanford.edu/entries/love/

[Revised entry by Bennett Helm on July 9, 2009.
Changes to: Main text, Bibliography]
This essay focuses on personal love, or the love of particular persons as such. Part of the philosophical task in understanding personal love is to distinguish the various kinds of personal love. For example, the way in which I love my wife is seemingly very different from the way I love my mother, my child, and my friend. This task has typically proceeded hand-in-hand with philosophical analyses...

http://www.shockmd.com/2009/07/09/the-synaptic-cleft-explained-in-raphip-hop/

http://www.shockmd.com/?p=3212

This parody of Wu-Tang Clan’s “Gravel Pit” was made for Human Biology 4A’s unit on Neuroscience. The unit is taught by Russ Fernald (who is featured on his bicycle).

For the lyrics and other background information check out SciVee

Thanks Mind Hacks

http://feedproxy.google.com/~r/scienceblogs/cognitivedaily/~3/gT4ld3QuhBA/smells_we_cant_detect_affect_j.php

http://scienceblogs.com/cognitivedaily/2009/07/smells_we_cant_detect_affect_j.php

[Originally posted in December, 2007]

Do smells have an impact on how we judge people? Certainly if someone smells bad, we may have a negative impression of the person. But what if the smell is so subtle we don't consciously notice it? Research results have been mixed, with some studies actually reporting that we like people more when in the presence of undetectable amounts of bad-smelling stuff. How could that be?

A team led by Wen Li believes that the judges might have actually been able to detect the odor, and then accounted for it in their response -- giving a face the benefit of the doubt when there's a hint of bad odor.

But odor detection is a tricky thing. Sometimes you're not sure if your milk or wine has gone bad, even after giving it a good whiff. The researchers felt that controlling the odors for a study would be the key to getting good results.

They first determined the odor detection threshold for each of 39 student volunteers. This was done by having each person sniff bottles containing progressively stronger solutions of three different compounds: Citral ("lemon"), anisole ("ethereal"), and valeric acid ("sweat"). The threshold was determined by when they could detect the odor. Then, for the actual experiment, bottles that were about 100 times more dilute were used.

Read the rest of this post... | Read the comments on this post...

http://www.chomsky.info/2009_07_01_archive.htm#8482932607706569259

New chomsky.info article

Season of Travesties: Freedom and Democracy in mid-2009. chomsky.info. July 9, 2009.

http://brainblogger.com/2009/07/09/natural-good-chemical-bad-right/

http://brainblogger.com/?p=3005

Arsenic sandwich anyone? Mercury soup, deadly nightshade surprise? No? Really? Well, I’m baffled! They’re all natural you know. And as we know, natural is good; natural is pure. Best of all, natural is healthy.

Such is the creed that has grown up around natural products. You want to market a new range of face cream –- make sure everyone knows it is natural. You want your expensive new yogurt to sell –- include the word “natural” on the packaging. The word “natural” has become byword for purity, health and goodness. … [visit site to read more]

---
Related Articles at Brain Blogger:

• Beat Depression with Brain Food while Eating Junk Food
• How Do We Feed Our Children?
• Defining Malpractice During an Emergency Evacuation
• Food Additives, Hyperactivity, and Common Sense
• The Trans Fat Ban – Is High-Fructose Corn Syrup Next?

http://www.stat.columbia.edu/~cook/movabletype/archives/2009/07/more_on_pearlru.html

To continue with our discussion (earlier entries 1, 2, and 3):

1. Pearl has mathematically proved the equivalence of Pearl's and Rubin's frameworks. At the same time, Pearl and Rubin recommend completely different approaches. For example, Rubin conditions on all information, whereas Pearl does not do so. In practice, the two approaches are much different. Accepting Pearl's mathematics (which I have no reason to doubt), this implies to me that Pearl's axioms do not quite apply to many of the settings that I'm interested in.

I think we've reached a stable point in this part of the discussion: we can all agree that Pearl's theorem is correct, and we can disagree as to whether its axioms and conditions apply to statistical modeling in the social and environmental sciences. I'd claim some authority on this latter point, given my extensive experience in this area--and of course, Rubin, Rosenbaum, etc., have further experience--but of course I have no problem with Pearl's methods being used on political science problems, and we can evaluate such applications one at a time.

2. Pearl and I have many interests in common, and we've each written two books that are relevant to this discussion. Unfortunately, I have not studied Pearl's books in detail and I doubt he's had the time to read my books in detail also. It takes a lot of work to understand someone else's framework, work that we don't necessarily want to do if we're already spending a lot of time and effort developing our own research programmes. It will probably be the job of future researchers to make the synthesis. (Yes, yes, I know that Pearl feels that he already has the synthesis, and that he's proved this to be the case, but Pearl's synthesis doesn't yet take me all the way to where I want to go, which is to do my applied work in social and environmental sciences.) I truly am open to the probability that everything I do can be usefully folded into Pearl's framework someday.

That said, I think Pearl is on shaky ground when he tries to say that Don Rubin or Paul Rosenbaum is making a major mistake in causal inference. If Pearl's mathematics implies that Rubin and Rosenbaum are making a mistake, then my first step would be to apply the syllogism the other way and see whether Pearl's assumptions are appropriate for the problem at hand.

3. I've discussed a poststratification example. As I discussed yesterday (see the first item here), a standard idea, both in survey sampling and causal inference, is to perform estimates conditional on background variables, and then average over the population distribution of the background variables to estimate the population average. Mathematically, p(theta) = sum_x p(theta|x)p(x). Or, if x is discrete and takes on only two values, p(theta) = (N_1 p(theta|x=1) + N_2 p(theta|x=2)) / (N_1 + N_2).

This has nothing at all to do with causal inference: it's straight Bayes.

Pearl thinks that if the separate components p(theta|x) are nonidentifiable, that you can't do this, and you should not include x in the analysis. He writes:

I [Pearl] would really like to see how a Bayesian method estimates the treatment effect in two subgroups where it is not identifiable, and then, by averaging the two results (with two huge posterior uncertainties) gets the correct average treatment effect, which is identifiable, hence has a narrow posterior uncertainly. . . . I have no doubt that it can be done by fine-tuned tweaking . . . But I am talking about doing it the honest way, as you described it: "the uncertainties in the two separate groups should cancel out when they're being combined to get the average treatment effect." If I recall my happy days as a Bayesian, the only operation allowed in combining uncertainties from two subgroups is taking a linear combination of the two, weighted by the (given) relative frequencies of the groups. But, I am willing to learn new methods.

I'm glad that Pearl is willing to learn new methods--so am I--but, no new methods are needed here! This is straightforward, simple Bayes. Rod Little has written a lot about these ideas. I wrote some papers on it in 1997 and 2004. Jeff Lax and Justin Phillips do it in their multilevel modeling and poststratification papers where, for the first, time, they get good state-by-state estimates of public opinion on gay rights issues. No "fine-tuned tweaking" required. You just set up the model and it all works out. If the likelihood provides little to no information on theta|x but it does provide good information on the marginal distribution of theta, then this will work out fine.

In practice, of course, nobody is going to control for x if we have no information on it. Bayesian poststratification really becomes useful in that it can put together different sources of partial information, such as data with small sample sizes in some cells, along with census data on population cell totals.

Please, please don't say "the correct thing to do is to ignore the subgroup identity." If you want to ignore some information, that's fine--in the context of the models you are using, it might even make sense. But Jeff and Justin and the rest of us use this additional information all the time, and we get a lot out of it. What we're doing is not incorrect at all. It's Bayesian inference. We set up a joint probability model and then work from it. If you want to criticize the probability model, that's fine. If you want to criticize the entire Bayesian edifice, then you'll have to go up against mountains of applied successes.

As I wrote earlier, you don't have to be a Bayesian (or, I could say, you don't have to be a Bayesian)--I have a great respect for the work of Hastie, Tibshirani, Robins, Rosenbaum, and many others who are developing methods outside the Bayesian framework)--but I think you're on thin ice if you want to try to claim that Bayesian analysis is "incorrect."

4. Jennifer and I and many others make the routine recommendation to exclude post-treatment variables from analysis. But, as both Pearl and Rubin have noted in different contexts, it can be a very good idea to include such variables--it's just not a good idea to include them as regression predictors.) If the only think you're allowed to do is regression (as in chapter 9 of ARM), then I think it's a good idea to exclude post-treatment predictors. If you're allowed more general models, then one can and should include them. I'm happy to have been corrected by both Pearl and Rubin on this one.

5. As I noted yesterday (see second-to-last item here), all statistical methods have holes. This is what motivates us to consider new conceptual frameworks as well as incremental improvements in the systems with which we are most familiar.

Summary . . . so far

I doubt this discussion is over yet, but I hope the above notes will settle some points. In particular:

- I accept (on authority of Pearl, Wasserman, etc.) that Pearl has proved the mathematical equivalence of his framework and Rubin's. This, along with Pearl's other claim that Rubin and Rosenbaum have made major blunders in applied causal inference (a claim that I doubt), leads me to believe that Pearl's axioms are in some way not appropriate to the sorts of problems that Rubin, Rosenbaum, and I work on: social and environmental problems that don't have clean mechanistic causation stories. Pearl believes his axioms do apply to these problems, but then again he doesn't have the extensive experience that Rosenbaum and Rubin have. So I think it's very reasonable to suppose that his axioms aren't quite appropriate here.

- Poststratification works just fine. It's straightforward Bayesian inference, nothing to do with causality at all.

- I have been sloppy when telling people not to include post-treatment variables. Both Rubin and Pearl, in their different ways, have been more precise about this.

- Much of this discussion is motivated by the fact, that, in practice, none of these methods currently solves all our applied problems in the way that we would like. I'm still struggling with various problems in descriptive/predictive modeling, and causation is even harder!

- Along with this, taste--that is, working with methods we're familiar with--matters. Any of these methods is only as good as the models we put into them, and we typically are better modelers when we use languages with which we're more familiar. (But not always. Sometimes it helps to liberate oneself, try something new, and break out of the implicit constraints we've been working on.)

http://feedproxy.google.com/~r/scienceblogs/wDAM/~3/0wohDy4YGko/the_runner.php

http://scienceblogs.com/cortex/2009/07/the_runner.php

John Branch has an absolutely fascinating and beautifully told article in the Times today on Diane Van Deren, one of the premier ultra-runners in the world. Last year, she won the Yukon Arctic Ultra 300, which follows the treacherous trail of the Yukon sled dog race for hundreds of miles. (She was the first woman to ever complete the 430 mile version of the race.) This weekend she's participating in a race in Colorado that has a total elevation gain of 33,000 feet. But here's the neuroscientific twist: Diane is missing a chunk of her right temporal lobe, which makes it easier for her to engage in such stunning feats of endurance:

Don Gerber, who works at Craig Hospital, a rehabilitation hospital in Englewood, Colo., for people with brain or spinal-cord injuries, said that Van Deren "can go hours and hours and have no idea how long it's been." Her mind carries little dread for how far she is from the finish. She does not track her pace, even in training. Her gauge is the sound of her feet on the trail.

"It's a kinesthetic melody that she hits," Gerber said. "And when she hits it, she knows she's running well."

Of course, such a timeless existence - Van Deren seems like a perfect example of flow - comes with some real costs. She also experiences severe memory problems:

Van Deren struggles to remember people she recently met and has missed flights simply by getting too involved in a conversation at the gate.

"She never remembers where she parked," Page said. "Never, not once, to this day."

The lapses are not always amusing. Her husband placed photo collages around the house to help his wife remember vacations and family milestones that slipped past her memory's reach. Robin Van Deren, the 21-year-old middle child, recently told her mother that she lost a part of her in the surgery. They cried together.

Brenda Milner (at least according to my neuroanatomy textbook) helped pioneer the study of right temporal lobe deficits. She emphasized the lack of visual memory, which is certainly apparent in Diane. But I'm most intrigued by the absence of time awareness - when Diane is running it's as if she stops thinking about the clock. Interestingly, such awareness seems to depend in large part on the right hemisphere.

Read the comments on this post...

http://www.shockmd.com/2009/07/09/changes-over-time-in-digital-literacy/

http://www.shockmd.com/?p=3135

Digital literacy is the ability to employ a wide range of cognitive and emotional skills in using digital technologies. 6 digital skills:

(a) Photovisual literacy is the ability to work effectively with digital environments, such as user interfaces, that employ graphical communication. (b) Reproduction literacy is the ability to create authentic,meaningful written and artwork by reproducing and manipulating preexisting digital text, visuals, and audio pieces. (c) Branching literacy is the ability to construct knowledge by a nonlinear navigation through knowledge domains, such as in the Internet and other hypermedia environments. (d) Information literacy is the ability to consume information critically and sort out false and biased information. (e) Socioemotional literacy is the ability to communicate effectively in online communication platforms such as discussion groups and chatrooms. (f ) Real-time thinking skill is the ability to process and evaluate large volumes of information in real time, such as in computer games and chatrooms

These skills were tested again after 5 years in 18 high school students, 16 college students and 17 adults. A control group of 60 participants of the same age and background were also tested. All age groups showed a significant increase in performance, the adult group improved significantly more than the younger participants. This almost closed the gap between these groups in digital literacy. Only in the creative skills the performance of the younger groups decreased significantly (information task), whereas the adults improved slightly. Because the control group performed similar as the experimental group this can not be attributed to age factor or retest. Five years of experience and training improved users’ performance with digital technologies with only negative consequences for creative skills especially in the young.

The tasks were as follows: (a) Photo-visual task: Create a theater stage, using an unfamiliar interactive multimedia computer program that utilizes a graphic user interface. (b) Reproduction task: Using Microsoft Word, modify the meaning of an existing seven-line paragraph (about 100 words) by rearranging sentences, words, and letters. (c) Branching task: Plan a one-week trip to a country that was not visited by the participants, using an Internet tourist site. (d) Information task: Write a comparative, critical report on a news event that was reported in a biased way in five different Internet news resources

Experience with technology, and not age, accounts for the observed lifelong changes in digital literacy skills. Worrisome is the finding that the younger group showed a decrease in creativity and critical thinking using digital technologies in contrast to the older group who improved on this skill. Important message for educators?

Eshet-Alkalai, Y., & Chajut, E. (2009). Changes Over Time in Digital Literacy CyberPsychology & Behavior DOI: 10.1089/cpb.2008.0264

http://www.stat.columbia.edu/~cook/movabletype/archives/2009/07/democrats_do_be.html

I visited AT&T Labs today--lots of fun, a great group of people, an interesting mix of statistics and machine learning. They showed me some cool visualizations that I'll display soon.

Anyway, while I was there, somebody asked me about voters with different educational levels. In discussing it, we realized we wanted to break this down by ethnicity and age. So I quickly prepared a grid of graphs for him.

On the train ride back, I spent a few minutes making the graphs prettier:

These are based on raw Pew data, reweighted to adjust for voter turnout by state, income, and ethnicity. No modeling of vote on age, education, and ethnicity. I think our future estimates based on the 9-way model will be better, but these are basically OK, I think. All but six of the dots in the graph are based on sample sizes greater than 30.