A logistic regression is similar to a discriminant function analysis in that it tells you the extent to which you can predict a given variable based on what you know about other categorical variables. However, the advantage of logistic regression is that any number of variables can be included, and if desired, all predictor variables may be categorical. In SPSS, you can graph a logistic regression through the "Options" menu of the "Binary logistic regression" window.
Start SPSS. Select "Open an existing data source" from the welcome window that appears. Double-click "More Files," then navigate to your data file. Double-click the file to open it in SPSS.
Video of the Day
Click "Analyze," then "Regression" and then select "Binary Logistic." The "Logistic Regression" window will appear.
Click your dependent variable from the list on the right -- that is, the variable you are trying to predict. Then, click the arrow next to the "Dependent" box. Next, select your predictor variables, using the "Ctrl" button if you need to click more than one, and click the arrow next to the "Covariates" box. Note that your categorical variables automatically receive a "(cat)" label next to them. If you wish to include the interaction between any of your variables in the analysis, click each once on the main list on the left, then click the ">a*b>" button next to the "Covariates" box.
Select "Forward: LR" from the "Method" drop-down menu. This gives the results for each of your predictors separately, allowing you to see how much each one contributes to the overall model, as well as the predictive power of all variables together.
Click "Options." From the "Statistics and Plots" header, select "Classification plots." After doing this, SPSS returns a graph of your logistic regression. Other useful statistics from this menu are "Hosmer-Lemeshow goodness-of-fit" and "Iteration history." The output of these two tests gives you information on how accurate the model is. Click "Continue" when you're done.
Click "OK." Wait a while for the results to appear. The logistic regression is a fairly computer-intensive procedure and, with large datasets, this may take some time. When the output screen appears, scroll down to see your graph.