Today a deal so unbelievable that it can’t be believed!
No, wait. In fact, the deal was so unbelievable that it couldn’t be believed. This follows logically: that which is unbelievable cannot be believed. And since I was anxious that readers believe the deal, I had to soften it from unbelievable to believable-but-shocking.
Can you believe this?
All new subscribers will receive a PDF copy of Chapter 10 of Uncertainty: The Soul of Modeling, Probability & Statistics (all current subscribers should have already received their copy a while back; if not, email me).
I hope you were sitting down when you read that. It was a believable deal, but it was also a shocking deal. Please don’t sue me if you became so excited that you lost control of your extremities and injured yourself or others. Please don’t sue anybody. Suing people is no fun. Besides, you were warned.
Head over to the Subscription page, fill in any amount to be deducted monthly from your credit card, making sure to use your correct email, and that’s it! You can cancel any time—not that you’d want to.
Running the blog isn’t free: there are server fees, backup storage fees, code fees, equipment fees, and so on. Remember: I’m wholly independent and not tied to any regular source of funding, which includes, which mostly includes, government grants. And beside these yearly appeals, there is no advertising on the site.
Except for this. Modify that last claim with this exclusion in mind.
What’s the money for? Blog expenses, as mentioned, plus, I’m saving up for a decent camera to graduate the podcasts to videos.
Reminder. It’s not necessary, and you don’t have to donate to get them, but you can have daily posts emailed to you. Go to the very bottom of the blog, on the left side, and input your email. Easier than remembering to show up regularly.
The Ideal Christmas Gift
Get the book everybody is talking about—when they talk about this book. Yes, none other than the award-eligible Uncertainty: The Soul of Modeling, Probability & Statistics.
Sure, you probably have your own copy now, but it’s a good bet your mother has been living without. Or your nephew. Or that guy down the gas station who always allows you free refills on your Stupendous Size pop. Or if you happy to know anybody who in any way, even remotely, calls themselves a scientist or researcher. They’ll all need their own copies. Don’t skimp for yourself, either. Pages wear out, you know. Buy often, and buy early.
Bonus Chapter 10 Opening!
Here are the opening paragraphs of Chapter 10 (with math and references suitably adapted).
“A genuine expert can always foretell a thing that is 500 years away easier than he can a thing that’s only 500 seconds off. ——Mark Twain
An entire book could be written of various implementations of models in the predictive, observable form Pr(Y in y | X,D,M)$ (see the previous Chapter for the explanation of this form). Here I can do no more than cover those areas that seem most important to decisions common in science. I emphasize not so much particular models by specific persons, but how model results should be communicated and the errors usual in the classical methods. Universally, statistical results are presented as if they were not conditional on a model, which of course all are. Over-certainty abounds.
Regression is of paramount importance. The horrors to thought and clear reasoning committed in its name are legion, a fact which is well known, e.g. [various authors]. But it’s more than bad regression: misunderstandings of the nature of evidence are everywhere, but that this is so is increasingly gaining attention; see among many [these fellows], and in the hot field of neuroscience [like these guys]. It’s bad enough in academia, but if any reader has experienced consulting in non-academic settings, in, for instance, marketing, you will realize the problems detailed below are trivial. From my many experiences I have been able to discover that ordinary people think statistics is something akin to magic. The discussion on how statistical “control” is not control in the Section on regression should be read by everybody.
Reification is the deadly sin of modelling. The model is not the territory, though this fictional land is unfortunately where many choose to live. When the data do not match a theory, it is often the data that is blamed for marring a beautiful model. Models should never take the place of actual data, though they often do, particularly in regression and time series. Risk is nearly always exaggerated. The fallacious belief that we can quantify the unquantifiable, especially human emotions, is responsible for scientism. Hayek, in his Nobel prize speech, cautioned against assuming that the data we have, which is often times the only data we have, must therefore, because of its availability, be causal. This is a form of availability fallacy. Incidentally, Hayek also recommended (a version of) the predictive approach, especially with economic data. “Smoothed” data is often given pride of place over actual observations. Over-certainty is, as I have already claimed, at pandemic levels.
The general, overarching admonition is to escape the Cult of the Parameter. Speak of observables and not parameters. Models should be used in the predictive sense and checked against reality.
Because this Chapter describes the some of the many (infinite?) ways probabilistic thinking can go awry, it is more conversational in tone. Finally, at the end, I express some hope about the future.