Number 1 Book In Epistemology & In Logic! Get It While It’s Hot!

r21b

It’s probably gone by now, but for a little while my new book, Uncertainty: The Soul of Modeling, Probability & Statistics, was “#1 New Release in Epistemology” then “#1 New Release in Logic.” And this before its official release date, which is 22 July.

Pre-orders (I’m hoping) or some bizarre statistical fluke (more likely) account for it (one sale was probably enough to do it). In case it is pre-orders, now’s your time to contribute to the effort towards Number One Overall.

When I checked yesterday morning, I was something like 380,000th overall, which is nothing to brag about, except to you, dear loyal readers. Tonight (for it is now Tuesday night around 8:30 PM) it’s up to 103,000. At this rate of increase I’ll beat even Bill Clinton’s new memoir I Slept With Who? Or Is It Whom?.

List price is $70, which isn’t so cheap—but it’s worth every farthing. Too, Amazon has a history of knocking a few bucks off the list price. And pre-orders have a “price guarantee“, so that if there is a discount, you’ll get it.

r22b

A reader reminded me I promised to put up Steve Goldberg’s Foreword, which I’ll do after I create a full-time book page (next week). I’ve also started on questions which can be used as homework, of which I’ll post some (all?) on line, in case any should like to use the book for a class.

Which somebody should. (It’d be me if somebody wants to hire me to do it.)

Besides showing What Probability Is and, more importantly, What Probability Is Not, and giving the relation of probability to logic and epistemology, I call for a complete and utter change in the practice of probability and statistics. Eliminate all hypothesis tests. All in all, whether Bayesian (Bayes factors) or frequentist (p-values). Eliminate all parameter estimation. Again, all in all, whether Bayes or frequentist. No confidence or credible intervals.

In its place? Pure probability.

This entails a recognition of the conditional nature of probability, and a firm understanding that it is not decision. Wait: that’s the big reveal, though it doesn’t seem like it. Just like a magician does the trick long before anybody knows to look for it, saying all probability is conditional is the solution to all the standard problems. (The method is, of course, not a panacea. Uncertainty remains.)

A call for the release of models-as-predictions, so that any result may be checked by anybody. Scientists don’t even have to release their data, or admit their methods, but they do have to make predictions—which they’ll live or die by.

You know. In just the way science used to be done.

With that in mind, there is an insistence for the return to an understanding of cause and essence. Probability depends on cause and essence. Which everybody used to know—recall the gambler’s fallacy? that implied knowledge of both cause and essence—but somehow forgot with the advent of easy modeling.

The standard “problems” in epistemology disappear in the pure probability approach. No more paradox. There’s nothing wrong with infinity, but it’s playing with a raging fire. We’ve lost the idea infinity-based methods—statistical parameters, analysis, and so on—are only useful as approximations because all measurement and all decision are discrete and finite. All as in all. Don’t use infinity unless you’ve understood it.

Don’t use simulations. Simulations are not magic. Since there is no such thing as randomness, there is no such thing as “random numbers”, so there is no such thing as “random numbers” “blessing” a simulation. No “variable” and no number “has” a distribution. You thus cannot “draw” from a “normal.”

What to use in the place of simulation? Control.

No, not the “control” people say they use in statistical models, which isn’t control at all, but is instead a bluff and a boast. Use real control, as in the kind they do in physical experiments.

Now this blog is polemical (mostly), but the book isn’t, except for portions of the last chapter, where I discuss the most common errors. The text is sobriety itself (though I have one lawyer and one cannibal joke). There is some hard going. I eschewed mathematics wherever I could, which isn’t always, and to discuss some philosophical positions requires concentration on the part of the reader.

There isn’t anything cultural in it, nor anything directly about global warming, though anybody who’s ever analyzed a time series (as in global warming) must read why they’re almost surely doing it wrong. This includes folks on both sides of the question.

Here are the Chapter Abstracts.

More to come!

Sorry no podcast this week. Finishing up class. Only a couple of days left. A dedicated book page is coming.

8 Comments

  1. I still find it fascinating that one has to write a book on this. In college, I was taught that the outcome of an experiment was conditional on what the input factors were and limits thereof. Extrapolation beyond those factors was done very cautiously, if at all. No fortune-telling, no “applies to the entire population even if it doesn’t, etc. Seems somehow the media and politics, or laziness on the part of professors, something, changed to “we can predict the future”. Maybe it was computer models. Maybe it was personal injury weasels—I don’t know, but it certainly changed.

  2. Bob Lince

    I’ve heard of “cause and effect” and “cause and event” but what is “cause and essence”?

  3. Bob Lince

    Sheri,

    I was hoping to get something a little more in the way of an executive summary than a 47 page pdf as an answer, but thanks for the pointer anyway.

    BTW, I lost a bet with myself. I laid odds that the first response to my question would be: buy the book.

  4. Bob: Sorry, but there really is no short answer. All the answers I can come up with are research papers or books. (“Buy the book” never really occurred to me—I must really be out of it 🙂 )

Leave a Reply

Your email address will not be published. Required fields are marked *