Stat Computing > SPSS > Code

SPSS Code Fragment
Graphing results in logistic regression

Say you run a logistic regression, and you would like to show a graph with the y axis having the probability of the event and the x axis being your predictor.  The following shows how you can construct such a graph.

Say that you do a logistic regression and the coefficients are
   Constant is -3
    x1 is.3
    x2 is .1

Also, say the mean of X2 is .5

Say that you want to make a graph of the probability of Y by X1 showing X1 from 1 to 30, and hold all other variables constant at their mean (i.e. X2 would be .5).  The following program generates data for X1 from 1 to 30, and computes the predicted value of Y for these values, and then makes a graph of the results.  

* create a new data file that just has "x1" in it from 1 to 20 by 1.
input program.
  loop #i = 1 to 30 by 1.
  compute x1 = #i.
  end case.
end loop.
end file.
end input program.
execute.

* fill in logistic equation below .
* say equation is -3 + .3*x1 + .1 * x2.
* and x2 has a mean of 5.
compute ylog = -3 + .3*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
execute.

* show graph with prob y on y axis, and x1 on x axis.
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

This produces the following graph.  

You can make a graph of your own by ...

1. Change "1 to 30 by 1" to the range of values you want for your predictor.
2. Change the "compute ylog =" to have the data from your regression.

Here are examples where we vary the intercept, using -10 then 5 then 0.

* say equation is -10 + .3*x1 + .1 * x2 (x2 has a mean of 5).
compute ylog = -10 + .3*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
execute.
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

compute ylog = -5 + .3*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
execute.
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

compute ylog = 0 + .3*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
execute.
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

Here are examples showing how the graphs change when the coefficient changes from .05 to .1 to .3 to .5.  

compute ylog = -5 + .05*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

compute ylog = -5 + .1*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

compute ylog = -5 + .3*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

compute ylog = -5 + .5*x1 + 5*.1  .
compute py = 1 - 1/(1 + exp(ylog)) .
GRAPH /LINE(SIMPLE)=VALUE( py ) BY x1 .

UCLA Researchers are invited to our Statistical Consulting Services
We recommend others to our list of Other Resources for Statistical Computing Help
These pages are Copyrighted (c) by UCLA Academic Technology Services