Herein our semi-annual reminder that the purpose of this blog is mercenary. Filthy lucre is both impetus and goal of its daily readings. Moola, the spondulicks, the government’s glory, the geetis, the green stuff. Lettuce pray. Let it flow.
The burden is on you, Dear Reader. Let’s make it happen. Spread the good news: The Dancing Briggs is for hire!
Yours Truly is wholly independent. Think of the benefits! He has no schedule; he has no affiliations; he has no employer, only transient masters. He has no office: No clock no cubicle no concerns. No office gossip, no endless HR-mandated reviews. No inefficiencies.
These are, of course, the exact same freedoms possessed by those fellows who collect pop cans for remuneration. The same opportunities for advancement and for bonuses, too. The same retirement plan. But the sidewalk walkers have one additional benefit I do not. They don’t have to advertise. I do.
Folks don’t generally understand the wild life we statisticians lead. Why, only yesterday I was deeply involved in measuring trans-rectal diameters for diagnosing constipation. And then there was the time at the Med School I grabbed a slippery black tube off the wall in Dr Oz’s office (this was before he became “Dr Oz” or was famous) and said, “What’s this?”
How was I to know he specialized in colonoscopies?
One too many digestive anecdotes. I shouldn’t have read the Pope’s latest interview. Or had all that pizza. Never mind!
Switch instead to urology. I was recently asked, given my well known antipathy, did I give clients, when they requested them, p-values?
I am a great sinner (in this and in many other things). I do.
I won’t cop a plea of hypocrisy, though. I never give a p-value unless my arm has been twisted, and never without a lecture that there are no good uses of p-values, and that they only cause harm, and that they do not mean what it is thought they mean, and that they cannot prove cause or association or “links”.
So why give them? Because non-statistician journal editors always know more statistics than I. Most of my clients are academics pushing papers, and so they must bend their conclusions in the directions insisted upon by editors. And those editors always advocate doing things as everybody else does things. I have been lectured more times than our dear President has wagged his finger at the nation by editors telling me “The best test in cases like this is this-and-such.” (Hypothesis tests belong on the same pile with p-values.) The charge that peer review enforces mediocrity is true.
My strategy, then, is to provide the error, and also give the better or right answer using the predictive approach. At worst, it’s an opportunity. At best, we can slip in something new. And we had just such a success last week. Science, they say, is self-correcting. Which is, when you think about it, an admission that science is often wrong. Skip it.
What can I do for you? Eliminate the massive, cancerous over-certainty which plagues the old methods. Chances are, you have fallen prey to the Deadly Sin of Reification, or used pseudo-quantification, or have been misled into thinking you have identified the cause of some observation. Chances are, you won’t know the true definition of “chances are.”
The old ways tell you to be sure or nearly sure you’ve made the right decision. But this is almost always an exaggeration. Your confidence in some contingent conclusion should be curtailed if you rely on p-values and hypothesis testing.
Funny thing, though. People like the over-certainty which accompanies classical statistics. Indeed, it is the main reason for the great respect these methods enjoy. They make research a breeze! Want a “definitive” answer to some question? “Submit” your data to some software and out spits “the” answer. Just like a magic 8-ball. And just about as accurate, too.
What can I do for you? Analyses galore. Lectures, admonitions, seminars, classes, tutoring. You name it, I do it, then you pay. Preferably using large quantities of cash. Gold doubloons are always welcomed.
There is also Uncertainty: The Soul of Modeling, Probability & Statistics for those who would rather do it on their own.