Professor: Investigate whether logistic regression is the best model for your analysis.
Student: Okay - I'll probit.
Larry Lesser
Professor: Investigate whether logistic regression is the best model for your analysis.
Student: Okay - I'll probit.
Larry Lesser
Q. Who publishes books on real estate experiments for statisticians?
A. Random House.
Q. Who publishes the correlation and regression books that statisticians are always talking about?
A. Pearson
Larry Lesser
Correlation doesn't imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing "look over there"
Randall P. Munroe (1984 - )
Avoid Linear extrapolation ... The turkey's first 1000 days are a seemingly unending succession of gradually improving circumstances confirmed by daily experience. What happens on Day 1001? Thanksgiving.
John E. Sener (1954 - )
The invalid assumption that correlation implies cause is probably among the two or three most serious and common errors of human reasoning.
Stephen Jay Gould (1941 - 2002)
I don't see the logic of rejecting data just because they seem incredible.
Sir Fred Hoyle (1915 - 2001)
Lyric copyright by Dennis Pearl
may sing to tune of "I Walk the Line" (Johnny Cash)
I keep a close watch on this scatterplot of mine
I keep applying regression all the time
I keep R-Squared to be the tie that binds
Because that's fine,
I use the line
I find it very, very easy to do least squares
I find another optimum and nobody cares
Yes, I'll admit I'm a fool for rms errors
Because that's fine,
I use the line
As sure as night is dark and day is light
I keep regression on my mind both day and night
And Gauss-Markov proves that it is right
Because that's fine,
I use the line
I check for patterns in my residual plots
I check again for any wayward dots
I could try splines, but I can't find the knots
Because that's fine,
I use the line
I keep close to y when I have this x of mine
I keep applying regression all the time
I keep R-Squared to be the tie that binds
Because that's fine,
I use the line
Lyric ©2005, 2006, 2009 by Lawrence Mark Lesser
may sing to tune of "Mexican Hat Dance" (traditional)
For (X, Y) data pairs, we call the Y's
The values observed. Now, let's fit a line!
For each X, the value of Y where on the line you would hit
Is known as a fitted value-- the value we say we predict.
And those fitted Y's always wear a hat:
A caret or circumflex are other names for that.
Subtracting the Y hat from Y is (vertical) error defined;
The sum of the squares of all these we want to minimize.
And that is all done by the line of best fit,
But first make sure you plot the points you'd like to fit!
And when you go plot all the scatter, do you see linear trend?
And does everything all look random for errors versus the fits?