12 Comments
User's avatar
Gwyneth's avatar

"My heart soars like a hawk." I shall christian you Old Lodge Skins.

Expand full comment
PJ London's avatar

Many of us, especially those in business, have to make decisions. We know that our "models" are imprecise, we know that tomorrow everything we based our data on may change, we know that people and their choices/decisions are fickle and irrational, yet we have to make a decision.

At some future date we will be called to account for those decisions. If all goes well we get praise and hopefully financial rewards. When it goes wrong we get canned.

We have only two defences.

1) All the data pointed to our decision being right and we performed more than sufficient due diligence.

2) We formed a committee.

Any person with intelligence hates committees more than they hate being canned.

Hence models

Expand full comment
Crixcyon's avatar

No parameters...but I still like that word.

Expand full comment
Paul Fischer's avatar

Introducing ProbDepends™: The Ultimate Probability Leak Protection!

Tired of those pesky statistical surprises leaking out? With ProbDepends™, our state-of-the-art P-Value Pads absorb up to 99.9% of unexpected outcomes because nothing ruins a hypothesis like a rogue 0.05! Slip into confidence, whether you’re crunching data or dodging a Bayesian curveball. Get yours today before the next confidence interval catches you off guard!

Expand full comment
Paul Fischer's avatar

An insightful lecture, Dr. William! It is remarkable to consider how much of the statistical theory we learned, foundational concepts included, relies on the notion of infinity, an abstract construct that is unattainable. Perusing Casella and Berger’s Statistical Inference, one encounters the Central Limit Theorem as a cornerstone, permeating the text, its development intricately tied to the elusive infinity. Every significance test and CI count on something unattainable.

Expand full comment
William M Briggs's avatar

It’s far from easy. I don’t think anybody alive today, that I know of, has it all figured out.

Expand full comment
Paul Fischer's avatar

That is for sure. Always loved it though. Pity most of it doesn't really work.

Expand full comment
Fabius Minarchus's avatar

Sure, using a normal distribution to model a discrete set can give some absurd values. But that's why we have the binomial distribution.

Expand full comment
William M Briggs's avatar

The binomial also has parameters. And arises from letting letting N → infinity in the example today.

Something better as mentioned last time would be something more like a parameterless multinnomial. Something that gives probability to just those values, and only those values, in the finite discrete set of possible GPAs.

Expand full comment
Lon Guyland's avatar

Exist. Hmmm.

Does the Mandelbrot set exist? What do we mean when we say the solution to a differential equation “exists”?

Not saying you’re wrong, but the concept of “exists”, I think, is quite deep and the word is not without ambiguity.

Ultimately (if you take my meaning), it would seem to me that we have to qualify “exists” with “relative to what?”. Can we actually say that any *thing* (or concept) has absolute existence?

Expand full comment
William M Briggs's avatar

Numbers don’t exist either, in the way I’m using the word. Have being.

Things exists, substance exist, and these have natures and powers. Probability has none.

Expand full comment
Lon Guyland's avatar

Yes, I understand, and thank you for permitting me my little tangent.

Expand full comment