Times change

A few years ago (more than a few, now that I think about it), a sociology professor told me that, back in the Sixties, Texas Instruments would accept any Ph.D. as qualification to work there, regardless of what area the degree was in. I never had a chance to ask someone from TI about this but I recently came across a kind of confirmation in a crime novel of all places.

The Thief Who Couldn't Sleep, was Lawrence Block's first attempt at a continuing character and, unlike his other series, this and the books that followed are now mainly well-written period pieces of the mid-Sixties. The books are an odd mix of Ambler and Fleming centered around Even Tanner, an eccentric radical-at-large who actively supports every fringe group he can find ranging from left-wing separatists to a society dedicated to returning the House of Stuart to the British Throne.

To support this unusual lifestyle, Tanner supports himself writing theses for anyone with a few hundred to spare (the story behind that career choice would take too long to recount here). Most of his clients are business types looking to build their resumes. As he puts it:

Industry considers a bachelor's degree indispensable, and, by a curious extension, regards master's and doctorates as a way of separating the men from the boys. I don't understand this. Why should a Ph.D. awarded for an extended essay on color symbolism in the poetry of Pushkin have anything to do with a man's competence to develop a sales promotion campaign for a manufacturer of ladies underwear?

It would be nice to have some numbers to back up the anecdotes, but can you imagine anyone today saying, even in a work of fiction, that the way to get ahead in business is to get a doctorate in Russian literature?

(for a point of comparison, check out this post from YoungFemaleScientist)

"Research is a joy"

I tried to shoehorn this quote into a post I'm working on. I couldn't make it fit but I couldn't bring myself to throw it away either. So here it is.

The source is Lawrence Block's early novel, The Thief who Couldn't Sleep. The narrator is (among other things) a professional thesis writer working on a dissertation about Turkish prosecution of Armenians.

It was pleasant work. Research is a joy, especially when one is not burdened by an excessive reverence for the truth. By inventing an occasional source and injecting an occasional spurious footnote, one softens the harsh curves in the royal road of scholarship.

Tyler Cowen hacks Robert Reich's site

How else to explain this passage?

Apple’s supposed sin was to tell software developers that if they want to make apps for iPhones and iPads they have to use Apple programming tools. No more outside tools (like Adobe’s Flash format) that can run on rival devices like Google’s Android phones and RIM’s BlackBerrys.

What’s wrong with that? Apple says it’s necessary to maintain quality. If consumers disagree they can buy platforms elsewhere. Apple was the world’s #3 smartphone supplier in 2009, with 16.2 percent of worldwide market share. RIM was #2, with 18.8 percent. Google isn’t exactly a wallflower. These and other firms are innovating like mad, as are tens of thousands of independent developers. If Apple’s decision reduces the number of future apps that can run on its products, Apple will suffer and presumably change its mind.

There are, of course, economists who believe that the market will find a way to punish anti-competitve practices but Reich certainly isn't one of them. And Apple, though unquestionably innovative, has always aggressively tried to suppress competition dating back to their we-stole-it-first suit against Windows.

Apple's goal isn't to limit the number of apps they can run; it's to limit the number of programmers supplying apps to other platforms. Admittedly, Google and Microsoft are big boys that can take care of themselves, but with the iPad, Apple has one hell of a first mover advantge. When someone manages to come up with a competitive product, they will have to launch it under the most adverse conditions possible.

I agree with Prof. Reich that regulation of big banks is more important, but that doesn't mean that Apple isn't breaking anti-trust laws.

Weekly dose of Judson

NYT's best science writer has a new column up. This time she's looking at the placebo effect.

Are oil companies rational actors?

There are some good pieces out on the economics of oil spills from Jonathan Chait and Ezra Klein, but the best of the bunch is by William Galston:

Second, the oil well now spewing large quantities of crude oil into the Gulf of Mexico lacked a remote-control acoustic shutoff switch used by rigs in Norway and Brazil as the last line of defense against underwater spills. There’s a story behind that. As the Journal reports, after a spill in 2000, the MMS issued a safety notice saying that such a back-up device is “an essential component of a deepwater drilling system.” The industry pushed back in 2001, citing alleged doubts about the capacity of this type of system to provide a reliable emergency backup. By 2003, government regulators decided that the matter needed more study after commissioning a report that offered another, more honest reason: “acoustic systems are not recommended because they tend to be very costly.” I guess that depends on what they’re compared to. The system costs about $500,000 per rig. BP is spending at least $5 million per day battling the spill, the well destroyed by the explosion is valued at $560 million, and estimated damages to fishing, tourism, and the environment already run into the billions.

There’s something else we know, something that suggests an explanation for this sequence of events. After the Bush administration took office, the MMS became a cesspool of corruption and conflicts of interest. In September 2008, Earl Devaney, Interior’s Inspector General, delivered a report to Secretary Dirk Kempthorne that has to be read to be believed. One section, headlined “A Culture of Ethical Failure,” documented the belief among numerous MMS staff that they were “exempt from the rules that govern all other employees of the Federal Government.” They adopted a “private sector approach to essentially everything they did.” This included “opting themselves out of the Ethics in Government Act.” On at least 135 occasions, they accepted gifts and gratuities from oil and gas companies with whom they worked. One of the employees even had a lucrative consulting arrangement with a firm doing business with the government. And in a laconic sentence that speaks volumes, the IG reported: “When confronted by our investigators, none of the employees involved displayed remorse.”

Was leaving out these shut off switches a rational decision? Between direct costs, lawsuits and PR, BP may take a ten figure hit here, but they spent a great deal of money and effort lobbying not to take a measure that, to the casual observer, would seem to have a positive expected value.

Is it possible that, even for huge corporations, the natural distaste for being told what to do is stronger than the desire to maximize utility?

Authorship

I liked today's post by DrugMonkey on authorship. One item that I think is worth considering is that, in complex and multi-author projects, it's actually really hard to allocate credit on a paper by paper basis. That's because credit can't be broken into pieces that precisely fit "First Author", "Senior Author", "Second Author", and "Middle Author".

I have had the good fortune to be involved with some really exceptional co-authors so I can't actually recall a signigicant authorship dispute. But I do think that this highlights the need to rotate people on large projects to do the best to approximate the level of credit that people should get on projects.

But it's not an easy exercise and it gets harder the fewer the number of papers and the more junior the group (as early career researchers are very sensitive to credit allocation when it comes to being evaluated for employment and such things).

"Markets Are Not Magic"

An old post from Mark Thoma that's worth another look.

"Then I came up with a great idea: I could use the fork to spear the food and the spoon to scoop it."

As many others have noted, there is nothing journalists like more than a standard narrative and when reporting on education, there is no narrative more standard than the legend of the good teacher/bad teacher. No matter how complex, every issue can be explained by a disapproving account of an inept instructor or, better yet, a breathless paean to an inspiring educator.

I've mentioned before (here and here), there are some serious issues that need to be addressed (but almost never are) when comparing performance of teachers. Less serious but more annoying is the reporters' wide-eyed amazement at common classroom techniques. Things like putting agendas on the board or calling on students by name without asking for volunteers (see here) or having students keep a journal and relate lessons to their own life (see any article on Erin Gruwell). Things that many or most teachers already do. Things that you're taught in your first education class. Things that have their own damned boxes on the evaluation forms for student teachers.

These techniques are very common and are generally good ideas. They are not, however, great innovations (with a handful of exceptions -- Polya comes to mind) and they will seldom have that big of an impact on a class (again with exceptions like Polya and possibly Saxon). Their absence or presence won't tell you that much and they are certainly nothing new.

"Shape of the earth -- opinions still differ"

I was reminded of Paul Krugman's parody of a New York Times headline when I came to this NYT headline:

"Despite Push, Success at Charter Schools Is Mixed By TRIP GABRIEL"

Followed a few paragraphs later by the money shot:

But for all their support and cultural cachet, the majority of the 5,000 or so charter schools nationwide appear to be no better, and in many cases worse, than local public schools when measured by achievement on standardized tests, according to experts citing years of research. Last year one of the most comprehensive studies, by researchers from Stanford University, found that fewer than one-fifth of charter schools nationally offered a better education than comparable local schools, almost half offered an equivalent education and more than a third, 37 percent, were “significantly worse.”

Although “charter schools have become a rallying cry for education reformers,” the report, by the Center for Research on Education Outcomes, warned, “this study reveals in unmistakable terms that, in the aggregate, charter students are not faring as well” as students in traditional schools.

As I mentioned before, there is reason to believe that this research is biased in favor of charter schools.

If you showed me test results for a new cholesterol-controlling drug in which 20% of the subjects had lower LDL levels than when they started taking the drugs, 51% stayed the same and 37% were "significantly worse," I don't think I would describe the results as 'mixed.'

But, of course, I'm not writing for the New York Times.

Bringing a whole new meaning to the term "Primary Investigator"

From Talking Points Memo:

[Virginia AG] Cuccinelli has launched an investigation into one of the climate scientists who was embarrassed by last year's Climate-Gate controversy -- and in doing so, he may be challenging long-held norms about academic freedom.

Last month, reports The Hook of Charlottesville, the AG requested "a sweeping swath of documents" from the University of Virginia, relating to the climate research work -- funded through state grants -- of Michael Mann.

Mann worked at the university from 1999 to 2005, and now runs Penn State's Earth System Science Center. If he were found to have manipulated data, Cuccinelli could seek to have the research money -- plus damages -- returned to the state.

It's not clear that there's much evidence of that, however. The climate-gate emails showed some scientists discussing ways to keep views skeptical of global warming out of peer-reviewed journals, among other things -- but they did not show outright fraud. Nor did they undermine the broad expert consensus that man-made warming is occurring and must be addressed.

Mann's work is currently being investigated by Penn State. In a recent USA Today story, he defended it, saying that though errors might exist, they were not fraudulent.

Undiagnosed diseases

One thought that I have often had about prescription claims databases is that we often can't do anything with missing data. If a patient in such a database has undiagnosed hypertension, for example, it's unclear how to handle it. This is in stark contrast to cohort studies where a missing blood pressure reading is a clear case of missing data and straightforward multiple imputation may do wonders.

So I wonder if Andrew Gelman's idea for count data could be adapted for this purpose?

Or would we be buried by an excessive number of assumptions that might be required to make it work in these settings?

Prelinking Nick Krafft

When I get my thoughts on the subject sorted out I plan to write some posts on maximum utility and fitness landscapes. When I do, I'll certainly want to link to this post I saw on Economist's View:

In which I “attack old-fashioned economics,” i.e. utility maximization, by Nick Krafft: At an off-campus discussion toward the end of my senior year of college, the topic of behavioral economics came up. Leading the discussion was a professor of mine, David Ruccio- whose blog I link to regularly- who argued that to really move forward with these iconoclast ideas, we still have to get rid of the max u thing- it’s holding everything back. I didn’t really agree with him at the time, or I just didn’t know, but a recent panel I attended helped clarify why Ruccio, and other heterodox economists before him, are right, even if the panelists themselves don’t want to see it it or admit.

Hypertension or Blood Pressure?

So which one do you use in your statistical models? Sometimes, in diagnosis based data sets, you don't have a choice (Hypertension is a diagnosis but blood pressure may not be captured).

It seems like a simple question but it includes a lot of complexity. The binary variable is well understood, known to be a relevant change in patient characteristics and can account for things like medication treatment. The continuous variable, whule it has a lot more information, needs some assumptions on spacification. For example, can we really assume linearity of an association between blood pressure and a clinical outcome? If we only have treated blood pressure is that the parameter of interest or is it the "underlying level of blood pressure"? If the later, we have a messy missing data problem.

I admit, as a statistics guy, I strongly incline towards the continuous version of the variable. But it is not at all clear to me that it is always the dominant choice for dealing with these types of varibles.

Landscapes and Lab Rats

In this post I discussed gradient searches and the two great curses of the gradient searcher, small local optima and long, circuitous paths. I also mentioned that by making small changes to the landscape being searched (in other words, perturbing it) we could sometimes (with luck) improve our search metrics without significantly changing the size and location of our optima.

The idea that you can use a search on one landscape to find the optima of a similar landscape is the assumption behind more than just perturbing. It is also the basis of all animal testing of treatments for humans. This brings genotype into the landscape discussion, but not in the way it's normally used.

In evolutionary terms, we look at an animal's genotype as a set of coordinates for a vast genetic landscape where 'height' (the fitness function) represents that animal's fitness. Every species is found on that landscape, each clustering around its own local maximum.

Genotype figures in our research landscape, but instead of being the landscape itself, it becomes part of the fitness function. Here's an overly simplified example that might clear things up:

Consider a combination of two drugs. If we use the dosage of each drug as an axis, this gives us something that looks a lot like our first example with drug A being north/south, drug B being east/west and the effect we're measuring being height. In other words, our fitness function has a domain of all points on our AB plane and a range corresponding to the effectiveness of that dosage. Since we expect genetics to affect the subjects react to the drugs, genotype has to be part of that fitness function. If we ran the test on lab rats we would expect a different result than if we tested it on humans but we would hope that the landscapes would be similar (or else there would be no point in using lab rats).

Scientists who use animal testing are acutely aware of the problems of going from one landscape to another. For each system studied, they have spent a great deal of time and effort looking for the test species that functions most like humans. The idea is that if you could find an animal with, say, a liver that functions almost exactly like a human liver, you could do most of your controlled studies of liver disease on that animal and only use humans for the final stages.

As sound and appealing as that idea is, there is another way of looking at this.

On a sufficiently high level with some important caveats, all research can be looked at as a set of gradient searches over a vast multidimensional landscape. With each study, researchers pick a point on the landscape, gather data in the region then use their findings to pick their findings and those of other researchers to pick their next point.

In this context, important similarities between landscapes fall into two distinct categories: those involving the positions and magnitudes of the optima; and those involving the search properties of the landscape. Every point on the landscape corresponds to four search values: a max; the number of steps it will take to reach that max; a min; and the number of steps it will take to reach that min. Since we usually want to go in one direction (let's say maximizing), we can generally reduce that to two values for each point, optima of interest and time to converge.

All of this leads us to an interesting and somewhat counterintuitive conclusion. When searching on one landscape to find the corresponding optimum of another, we are vitally interested in seeing a high degree of correlation between the size and location of the optima but given that similarity between optima, similarity in search statistics is at best unimportant and at worst a serious problem.

The whole point of repeated perturbing then searching of a landscape is to produce a wide range of search statistics. Since we're only keeping the best one, the more variability the better. (Best here would generally be the one where the global optimum is associated with the largest region though time to converge can also be important.)

In animal testing, changing your population of test subjects perturbs the research landscape. So what? How does thinking of research using different test animals change the way that we might approach research? I'll suggest a few possibilities in my next post on the subject.

A good Bayesian Textbook?

Say that one wanted to teach Pharmacoepidemiology students about Bayesian statistics. Say further that it was important that the book be clear and easy to follow. Are there any alternatives to Gelman and Hill (which is clear but remarkably free of drug related examples)?

Just wondering . . .