John Myles White: Die Sudelbücher

Lately I’ve been thinking about the optimal strategy for data collection when you plan to run a linear regression. Clearly, you want a sample of widely distributed points if you’re unsure that a strict linearity assumption is appropriate. If you already know from theoretical reasons that linearity is appropriate, then you know that you only need two correct (x, y) data points to uniquely define the regression line. To get this, one conventionally samples many (x, y) pairs and then computes the regression line’s slope and intercept. Why not sample only two x data points over and over again instead? If you are trying to find the formula for the line E[y | x], it seems reasonable to assume that high quality estimates of the points (a, E[y | x = a]) and (b, E[y | x = b]) would be a good way to do this.

Is this a reasonable approach to two variable linear regressions? Is this approach less efficient statistically than sampling at many points? Or is the reason to avoid this strategy in practice is that one is uncertain of the validity of the linearity assumption in all but exceptional cases?

Responsibility, in the moral sense, is not a property of actors, but merely a confabulation on our part designed for one and only one purpose: to justify the suffering we inflict upon others. This holds equally well in regard to matters of retributive and distributive justice: responsibility is a mere holdover from an antiquated conception of the world.

It is not responsibility, but the estimation of conditional recidivism that should concern any legal system.

I find extraordinary the readiness with which members of certain branches of the social sciences latch onto obscure and irrelevant findings in the natural sciences as means of buttressing their views. This habit convinces me that their work is, as Hayek would say, merely scientistic — that their life’s work is the application of a whitewash of scientific terminology to what is merely an exposition of their moral convictions.

It occurred to me today that the reason I find attempts to “beat Vegas” so odd is that I am nearly certain that, if you are smart enough to actually beat Vegas and yet you are not spending your time trying to beat Wall Street instead, you must be stupid.

Consider this my proof by contradiction that no one ever beats Vegas.

Among those with any compassion for the poor, there seem to be two schools: those who believe that the poor face difficult lives that make it impossible for them to function and those who believe that the poor face easy lives, but are constitutionally incapable of meeting even those limited challenges for some reason.

It occurred to me today that there should be a website devoted to web design called “Status R-Quo”.

Every time I have doubts about how safe flying is, I think to myself:

It is not from the benevolence of the pilot that I expect not to die after crashing into a mountain, but from the pilot’s regard to his own interest.

It amazes me how this line of reasoning seems to never enter into the minds of those whose fear of flying is so great that they simply refuse to board planes. It is as if they genuinely believe that the pilot, who flies many times each week, is not concerned with keeping himself alive.

I suspect the only relevant criterion for selecting potential pilots is an estimate of their disposition to suicidal behavior.

An 2005 article by Derek Sivers, entitled “Ideas Are Just a Multiplier of Execution,” has been making its way through programming blogs lately. In regard to software design, I think the article is very accurate: ideas are worth very little and execution is enormously more important. The most profitable software in the world is not especially inventive: consider Microsoft Windows, websites like Myspace or Facebook, or applications like iTunes and Microsoft Word. These bits of software bring in enormous profits each year without, in my opinion, coming close to the sort of non-obviousness that forms the basis of most modern nations’ patent systems.

This is not to say that you cannot find remarkably clever ideas forming basis for some software. Bayesian spam filtering, Google’s PageRank algorithm, and, above all else, the original conception of public key cryptography are ideas of enormous value. The value of the software implementing these ideas is much less based on the implementation’s quality as software than upon the ideas themselves. These ideas are ground-breaking: cryptography has been important to humans for millenia, yet only in the past fifty years has anyone been brilliant enough to realize that some ideas from 17th century developments in number theory allow us to create secret codes whose means of encryption can be made public knowledge without disclosing the means of decryption.

And once we leave the world of software design, the value of ideas increases several fold. Consider some of the greatest ideas of the past two centuries: the marginal revolution in economics, Mendel’s allelic theory of genetics, or Einstein’s replacement of absolute space and absolute time with the absolute speed of light.

These ideas are gold in and of themselves. Once you have Einstein’s profoundly counter-intuitive notion in mind, it takes a much less brilliant physicist to do the work that develops the special theory of relativity. Many physicists are capable of performing the Kuhnean “normal science” that brings relativity to fruition; Einstein alone seems to have had the ability to see when one was required to question and rethink the most basic assumptions of his subject.

In light of these great achievements and the wealth and good fortune they have given all of us, I want to praise ideas for their own sake. While I agree that we should not soon cease to remind ourselves of the central importance of execution — bearing in mind Thomas Edison’s oft-quoted quip that “genius is one percent inspiration and ninety-nine percent perspiration” — we should also remember that the value of ideas is non-trivial when the ideas themselves are non-trivial.

What strikes me as so absurd about the word progressive and its use when speaking of, say “progressive Denmark”, is the arrogant presumption that we know what events represent progress.

There are two recurrent complaints about American democracy that have been on my mind lately: the inability of most Americans to name their representatives and the ever-increasingly central role of the president. Both of these strike me as the inevitable result of our government’s structure. There simply are too many representatives to keep track of. Americans may be able to keep track of some five hundred celebrities, but this store of information tends only to include primordial human information: physical appearance, one’s sense of liking the person, etc.. This is plainly not the data one needs to vote intelligently.

I see the centrality of the president as the reaction to the overwhelming number of Congress members. It is possible to focus on a single point from which one’s other opinions derive. I would assume that statistics would support the assertion that most voters choose the Congress candidates of the party of their most recent presidential pick. The existence of a monarchical president simply answers basic human needs to simplify situations and make them comparable to the ancient tribal situations for which the human mind is optimized.