Running the Analysis

In order to properly examine the structure of the course of illness for each diagnosis, you will want to make the results of the projected centroids table available as data for scatterplots. You can accomplish this using the Output Management System.

1. To begin an OMS request, from the menus choose:

   Utilities > OMS Control Panel...

   

2. In the Output Management System Control Panel, select Tables in the Output Types list.
3. Select CATPCA in the Command Identifiers list.
4. Select Projected Centroids in the Table Subtypes for Selected Commands list.
5. Select File in the Output Destinations group and type projected_centroids.sav as the filename.
6. Click Options.

   

7. In the OMS Options dialog, select SPSS® Statistics Data File in the Format drop-down list.
8. Type TableNumber_1 in the Table Number Variable field.
9. Click Continue.
10. Click Add in the Output Management System Control Panel.
11. Click OK, and then click OK to confirm the OMS session.

    The Output Management System is now set to write the results of the projected centroids table to the file projected_centroids.sav.

12. To produce categorical principal components output for this dataset, from the menus choose:

    Analyze > Dimension Reduction > Optimal Scaling...

    

13. In the Optimal Scaling dialog, select Some variable(s) not multiple nominal in the Optimal Scaling Level group.
14. Click Define.

    

15. Select Body weight through Body perception (when variables are listed in file order) as analysis variables.
16. Click Define Scale and Weight.

    

17. In the Define Scale and Weight dialog, select Ordinal as the optimal scaling level.
18. Click Continue.
19. Select Time/diagnosis interaction as a supplementary variable and click Define Scale in the Categorical Principal Components dialog.

    

20. In the Define Scale dialog, select Multiple nominal as the optimal scaling level.
21. Click Continue.

    

22. In the Categorical Principal Components dialog, select Time of interview through Patient number (when variables are listed in file order) as labeling variables.
23. Click Options.

    

24. In the Options dialog, select Variable names or values in the Label Plots By group.
25. Click Continue.
26. Click Output in the Categorical Principal Components dialog.

    

27. Select Object scores in the Tables group.
28. Request category quantifications for tidi.
29. Choose to include categories of time, diag, and number.
30. Click Continue.
31. Click Save in the Categorical Principal Components dialog.

    

32. In the Save dialog, select Save to the active dataset in the Transformed Variables group.
33. Click Continue.
34. Click Object in the Categorical Principal Components dialog.

    

35. In the Object and Variable Plots dialog, choose to label by Variable in the Label Objects group, and then select time and diag as the variables to label objects by.
36. Click Continue.
37. Click Category in the Categorical Principal Components dialog.

    

38. Request category plots for tidi.
39. Request transformation plots for weight through body (when variables are listed in file order).
40. Choose to project centroids of tidi onto binge, satt, and preo.
41. Click Continue.
42. Click OK in the Categorical Principal Components dialog.

    The procedure results in scores for the subjects (with mean 0 and unit variance) and quantifications of the categories that maximize the mean squared correlation of the subject scores and the transformed variables. In the present analysis, the category quantifications were constrained to reflect the ordinal information.

    Finally, to write the projected centroids table information to projected_centroids.sav, you need to end the OMS request. Recall the OMS Control Panel.

43. In the Output Management System Control Panel, click End.
44. Click OK, and then click OK to confirm.

Next