2
$\begingroup$

I have a newbie question about logistic regression fit plots. I'm fitting a very simple binary output based on a simple continuous input

X   Y
0.1 0
0.1 0
0.1 0
0.1 1
0.5 0
0.5 0
0.5 1
0.5 1
0.9 0
0.9 1
0.9 1
0.9 1

when using JMP, the fitting model is logistic. I understand the fitted line, but what are the points plotted on the chart? I only have 0,1 in the output, but why does the logistic plot show values of Y that are not 0,1?

logistic scatter plot

Thanks for any help with this.

$\endgroup$

2 Answers 2

4
$\begingroup$

The plot that JMP uses for logistic fits is described on page 2-79 (page 93 of the PDF) in this demo document produced by JMP.

http://www.jmp.com/training/scripting_workshop/2004/concepts_using_jmp_handout.pdf

This platform features a new kind of scatter plot. The data points are plotted according to their real abscissa and a dummy random ordinate. It organizes the points so that you can see how they fall into one of the categories.

So for each point, the $X$ coordinate is the position that is in the data, and the $Y$ coordinate is just randomly placed above or below the blue line so you can see where the points fall.

$\endgroup$
2
  • $\begingroup$ Thanks for the explanation but what is the purpose of plotting a random value as the y axis? Is it completely random or some weighed random number based on the fitted model? $\endgroup$
    – Jonjilla
    Sep 28, 2014 at 5:44
  • $\begingroup$ @Jonjilla The point is probably to just jitter the data so that you can get a sense of where the data lies relative to the logistic curve. It doesn't look completely random, but the JMP guide still uses the word "random." $\endgroup$ Sep 28, 2014 at 21:04
0
$\begingroup$

The left y-axis location of these dots is meaningless as the location is completed random within the region above or under the curve. Also, the left y-axis reading is NOT the fitted probability of Y=1, rather it is of Y=0! What a confusing plot!

$\endgroup$
1
  • $\begingroup$ Think it might be worth emphasising, for clarity, that the left y-axis refers to the blue curve. I was also surprised by the fact this seems to show P(Y=0)! $\endgroup$
    – Silverfish
    Jul 2, 2016 at 19:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.