Please wait while the transcript is being prepared...
The clinical actuarial controversy relates to whether it's better
to make decisions using our own best human judgment,
based on our experience and training and so on,
or whether it's better to make decisions using statistical models.
These are called actuarial models.
They might be developed by multiple regression or analysis of variance.
The question is, which of these is better?
This is an ongoing controversy and with people
on both sides of the controversy have very strong feelings.
This becomes very emotional.
With the actuarial approach to judgment,
we simply eliminate the human judge.
Instead, we use forums such as multiple regression analysis to
find how predictor variables seem to combine and how they're weighted,
in order to best predict some criterion variables,
some outcome or event of interest.
For example, if we wanted to predict psychosis,
we might consider a set of variables that we think might be predictors
at the extent to which those variables do in fact predict psychosis,
and the resulting model, the actuarial model,
could then be used to predict whether
particular individuals would show signs of psychosis.
One interesting example of application of an actuarial model is
in prediction of the outcomes of Red Bordeaux Wine Auctions.
This is a case where red wines are young,
we'd like to know what they're going to sell for in auction.
The traditional approach is to have tasters clinically make judgments,
they swirl and smell and taste the young wine.
Ashenfelter and colleagues tried to predict
the quality of the vintage red Bordeaux wine statistically.
Taking the market price at auction of mature Bordeaux wines as their index of quality,
they showed how the advantage of the wine strongly determines its quality.
By the time of Bordeaux wine is mature and drinkable,
there's usually considerable agreement
among wine drinkers as to the quality of the vintage.
The trick, though, is to predict that quality decades
in advance when the wine is young and undrinkable.
Ashenfelter and colleagues observed that the weather during
the growing season is the key determinant of the quality of grapes.
Specifically they said, "Great vintages of
Bordeaux wines correspond to the years in which August and September are dry,
growing season is warm,
and the previous winter has been wet."
They developed a multiple regression model that is an actuarial model,
which they refer to as the Bordeaux Index.
That index considers the age of the vintage and indices reflect
the temperature and precipitation variables that we mentioned before.
Remarkably, that index accurately predicts 80 percent
of the variants and the price that mature Bordeaux red wine commands at auction.
This is much better than any clinical judgment.
Further, of course, the actuarial model makes
these predictions as soon as the growing season is over,
before any experts have even tasted the young wine.
Observers of the industry said the reaction of wine industry individuals to
the Ashenfelter red wine model was somewhere between violent and hysterical.
Nevertheless, the model gives wine drinkers and buying buyers a pretty useful tool.