Here are some quotations I like.
Part of the appeal of quotes and aphorisms is that they're concise, but a problem with conciseness is that it tends to create ambiguity. So, I've included after each quotation a description of what I think should be learned from it. Admittedly, these interpretations may not be the ones intended by the authors.
It don't matter if you're ugly in this racket [baseball]. All you have to do is hit the ball and I never saw anybody hit one with his face.
Apart from concerns about social norms, personal appearance doesn't matter. In general, appearance is overvalued.
A million guys can dunk a basketball in jail; should they be role models?
Skill is not the same thing as virtue, and the skills that bring fame and fortune are not always the skills that are most useful to society. In particular, the popular worship of sports (as opposed to, for example, playing video games, which is looked down upon instead of honored) as something intrinsically good is illegitimate. Sports are only games.
…the realization that radio and television are primarily media of entertainment was slow in coming, and the acknowledgment of this fact continues to disillusion numerous irrepressible media idealists…
Radio and TV can be used for informative or educational ends, but by the nature of broadcast, they're best suited to entertainment. The most successful educational TV shows, like Sesame Street, have generally blended education with entertainment. Audio or video with a more purely practical purpose is better distributed on discs or over the Internet, where it can be consulted at will.
I generally avoid temptation, unless I can't resist it.
This short, confusing sentence feels representative of the deep ambivalence inherent in self-control dilemmas (whether they're about sex or not). Self-control dilemmas are characterized by how we in some sense want to fail. The quote also suggests that we may semi-voluntarily leverage the very idea of lack of self-control (being "unable to resist" something) in order to justify an indulgence to ourselves and thereby enable it.
Protect me from what I want.
Desire and pleasure can be dangerous. More generally, we don't always act in our own best interests.
There is no wrong way to fantasize.
We can't be held accountable for our thoughts and desires. We have little to no control over them, and besides, they only matter in how they influence our actions, and our actions are what we should answer for.
God often gives nuts to toothless people.
There are no guarantees that resources, aptitude, and ambition are distributed among humanity in a convenient way. So, they tend to be hopelessly scattered. This is one way in which the world is inherently unjust.
In all of mankind's history, there has never been more damage done than by people who "thought they were doing the right thing".
As a rule, people have good intentions. Thus, most of the harm in the world is done by well-intentioned people. This phenomenon is related to Hanlon's razor: "Never attribute to malice that which is adequately explained by stupidity."
People are always frustrating us by interfering with the relationship we are trying to have with them.
People sometimes say they like a person just for who they are, but this isn't so, nor should it be. We like people not for merely existing, but for having particular qualities. Sometimes we go so far as to idealize other people, putting them on a pedestal. When the other person changes, or when we merely learn more about them, our concept of them and hence our feelings towards them may be challenged. A rigid or hyperbolically idealistic view of another person is especially vulnerable to such challenges and thus is not conducive to a stable relationship. On the flipside, if a person changes dramatically, we shouldn't cling too tightly to our old attitude towards them.
The problem with being consistent is that there are lots of ways to be consistent, and they're all inconsistent with each other.
One problem with conventions that's simple but seems rarely appreciated is that often, several conventions apply to the same situation, and they disagree. Sometimes you have to choose which convention to follow; sometimes it makes more sense to depart from all of the conventions. Aim for the final work to be coherent and useful.
…good synthesis is different from good compromise…
A decision that makes equal concessions to all interested parties (that is, a good compromise) is not necessarily a good decision. The end product (the quality of the synthesis) is more important than nominal fairness.
Beware of bugs in the above code; I have only proved it correct, not tried it.
Formal proof makes stronger guarantees than empirical testing, but these tend not to be the guarantees we want in applications. The best way to check whether a method works is to actually try it. Whether it works in practice is what matters; the theory is ultimately irrelevant.
Classification of mathematical problems as linear and nonlinear is like classification of the universe as bananas and non-bananas.
This one is a tad glib. There's a good reason that quantitatively minded people prefer linear problems: they're a lot more tractable. But don't be fooled into thinking that linear problems are inherently more common or more important.
Everyone believes this [that errors are normally distributed], Mr. Lippman told me one day, because the experimenters believe that it is a mathematical theorem, and the mathematicians that it is an experimentally determined fact.
The normal distribution is another case of mathematical elegance being conflated with natural law. It's hard to say how applicable asymptotic theory is in practice, and the celebrated ubiquity of the normal distribution in nature is due at least in part to how eager we are to use it.
After a hard day's work doing Data Science I like to come home and wind down with some Food Cooking and Word Reading before bed.
"Data science" is a silly term, because all science is about data. Recognizing this, by the way, makes it clear why all scientists should take data analysis seriously.
Politicians use statistics in the same way that a drunk uses lamp-posts—for support rather than illumination.
I've often seen statistics motivated as a rhetorical tool. But if statistics were useful only for convincing people of things, it would be little better than a course in persuasive writing. The power and utility of statistics (and more generally, of data analysis) is as a tool for discovery and understanding.
Incidentally, I've sometimes seen statistics extolled for its objectivity. Actually, statistics is highly subjective, at least if it's applied thoughtfully and well instead of automatically and carelessly. Several competent data analysts, given the same data and question to be answered, could easily arrive at different conclusions—by thinking about the problem in different ways, by making different assumptions, by optimizing different metrics of the quality of their analysis, and so on. This is not a problem so long as we evaluate data analyses on a case-by-case basis rather than insisting on certain rules (such as "always use t-tests") and then uncritically accepting the conclusions of analyses so long as they follow the rules.
We had an office pool [for a lottery], but I chose not to contribute on the argument that my $1 could buy me 1/6 of a burrito. And I went home that night and I bought a burrito. God help me, 1/6 of it tasted better than losing the lottery I can tell you that.
Playing the lottery is dumb. You will lose.
Every statistician, from R. A. Fisher on down, will use Bayesian methods when they are appropriate—that is, when there is a physical sampling distribution for the parameters. A Bayesian is someone who uses Bayesian methods even when they're inappropriate.
This is not an attack on Bayesians; Gelman is a Bayesian par excellence. It is a reminder that even for analysts who reject the Bayesian creed, Bayes's theorem and related tools are useful, even optimal, for solving some problems. Generally, when people say a person or a method is "Bayesian", they mean not merely that it employs Bayes's theorem but that it assigns a prior distribution when no such distribution already exists in the definition of the problem.