For students to appreciate why line of best fit is generally implemented by minimizing the sum of the squares of the errors, JSE Editor Michelle Everson suggested that The University of Texas at El Paso’s Lawrence Lesser write a parody of “Beat It,” Michael Jackson’s 1983 number-one hit that went platinum, won a Grammy, and helped make Thriller the best-selling album of all time. Here is Lesser’s result, which incorporates a couple of suggestions from Ethan Brown.

Square It!

They told him, don’t just add the regression errors here:
The plus and minus cancel and they disappear.
So make each error positive, make it really clear—
Just square it, just square it!
Absolute value doesn’t make me a fan:
A line it gives is not unique, ya understand?
And that is real tough for a software command,
So square it, you want to so bad:

Just square it (square it), square it (square it):
Get a better line of best fit!
It’s the least you can do to get the best line,
It really matters how you’re inclined:
Just square it (square it), just square it (square it),
Just square it (square it), just square it (square it)

Oooh! Markov and Gauss, you know that each was the man,
Showed squaring errors made a better plan.
They talked ’til they were BLUE ’bout their theorem so grand,
So square it, just square it.
Oooh! Well now we know to minimize residuals squared
‘Cause other estimators really can’t compare:
The coefficients’ estimates are most precise there,
So square it, you want to so bad:

Just square it (square it), square it (square it):
Now derivatives can bear it!
It’s the least you can do to get the best line,
It really matters how you’re inclined:
Just square it (square it), just square it (square it),
Just square it (square it), just square it (square it)

Lyrics © 2013 Lawrence M. Lesser (reprinted with permission)