Correlation/Regression

• What's the Method You Want?

lyrics by Greg Crowther and Leila Zelnick
may be sung to the tune of "You're the One That I Want" by John Farrar

DZ: You’ve got data from a trial
With a treatment and control;
An ANOVA can determine
If the treatment's workin'!

SO: You better shape up --
No ANOVA, man!
Use a tool that’s more precise!
You better shape up --
Do ANCOVA, man!
Make your error bars look nice!

DZ: Make 'em nice!

BOTH: CHORUS:
What's the method you want?
(Method that you want, ooh, ooh, ooh!)
It depends on the study you’ve got!
(Study that you've got, ooh, ooh, ooh, Honey!)
What's the method you want?
Method that you want, ooh, ooh, ooh!)
It just depends (just depends)!
Be smart, my friends (yes, my friends)!

SO: If your groups aren’t random,
Different baselines may be seen.
An ANCOVA will regress them
To the mean.

DZ: You better shape up,
’Cause you need to know...
SO: ...You need to know
How group changes should be scored.
That is unconditional...
SO: ...Unconditional,
Needs ANOVA, nothing more.
DZ: Nothing more...
BOTH: ...As explained by Frederic Lord!

BOTH: CHORUS

• The Conviction of Miss Prediction

by Dane C. Joseph

It began when an intercept
directed some concepts—
strength and direction—
to capture covariance.

A line mapped two variables,
a plane modeled more.
All seeking least-squares variance
and following the law.

But they ran into trouble
when a categorical response
tricked moderator and predictor
into romantic nonchalance.

By then it was too late
as in hindsight you could see
Linear r^2 was incapable
to do a Nagelkerke.

And I’m still not quite sure
why the sigs from the F’s,
were not up to snuff
as the Wald difference tests.

But in the courtroom we heard
the lawyers and judge
decry her dereliction...
then subpoenaed probability and odds.

• Spurious Correlation Sestina

by Jules Nyquist

when zero
is a temperature
it is an interval scale
that dips below an imaginary
line to go negative, as in a thermometer
in a pandemic used as permission to measure

our temperature, but how do we measure
something that falls below zero
like the weight of a bird—our thermometer
won’t register an invisible temperature
and we will disappear like the imaginary
checkbook balance of youth, on a scale

of probability, the chances of converting the Fahrenheit scale
to Centigrade was remote, the U.S. still measured
in Fahrenheit behind the rest of the imaginary
world where a bank balance waits to zero-
out and a raven pulls shiny coins from the sky in a temperature-
controlled out-of-the ether mainframe thermometer

six thousand feet above sea level, a thermometer
measures the speed on a speedometer scale
when the motorcycle driver hurls over a temperature-
reduced mountain ridge to an almost measurable
crash that soars into a stock-market zero-
point of ratio scaling, a lie on an imaginary

boundary where the motorcycle driver imagines
they never hit the car and the thermometer
never registered above human normal and zero
meant nothing, it was only an innocent bathroom scale
that we blamed added ten immeasurable
pandemic pounds to our weight and the temperature

of Earth rose only in the height of trees,
a temperature that didn’t take into account the imaginary
altitude sickness that turned out to be very measurable
and tripled the effect of the beer stored in a thermal
cooler, found by the side of the scaled
curve of that mountain road where zero

was just a measurement of temperature
and the imaginary paper bank statement never showed zero
due to there was nothing to scale on the erratic thermometer.

• My Data Lies Over the Average

may sing to the tune of "My Bonnie Lies Over the Ocean" (Scottish folk song)

My Data lies over the average
Much higher than I’ve ever seen
And if I take one more sample
Oh, expect it back towards the mean

My Data lies over the average
Much higher than ever seen
Well, give me one more sample
Yeah, expect it back closer to mean

Yeah bring back, ah bring back
Oh bring data more towards the mean, the mean
(Ah bring) Oh bring back, ah bring back
Oh bring back my Data to mean.

Well, my Data lies over the average
My proportion’s well over the p
Yeah, give me another proportion
Oh, expect it back closer to p

Yeah bring back, ah bring back
Oh bring back the data to mean to mean (to mean)
Oh bring back, ah bring back
Oh bring back my Data to mean.