Differential Development for Selected Variables

One variable from each bundle of symptoms identified by the component loadings was selected as “representative” of the bundle. Binge eating was selected from the bulimic bundle; sexual attitude, from the anorectic/psychosocial bundle; and body preoccupation, from the third bundle.

In order to examine the possible differential courses of illness, the projections of Time/diagnosis interaction on Binge eating, Sexual attitude, and Preoccupation with food and weight were computed and plotted in the following figure.

Figure 1. Projected centroids of Time/diagnosis interaction on Binge eating, Sexual attitude, and Preoccupation with food and weight
Projected centroids of Time/diagnosis interaction on Binge eating, Sexual attitude, and Preoccupation with food and weight

This plot shows that at the first time point, the symptom binge eating separates bulimic patients (2 and 3) from others (1 and 4); sexual attitude separates anorectic and atypical patients (1 and 4) from others (2 and 3); and body preoccupation does not really separate the patients. In many applications, this plot would be sufficient to describe the relationship between the symptoms and diagnosis, but because of the complication of multiple time points, the picture becomes muddled.

In order to view these projections over time, you need to be able to plot the contents of the projected centroids table. This is made possible by the OMS request that saved this information to projected_centroids.sav.

Figure 2. Projected_centroids.sav
Data view of the Data Editor show the contents of the data file Projected_centroids.sav

The variables Bingeeating, Sexualattitude, and Preoccupationwithfoodandweight in projected_centroids.sav contain the values of the centroids projected on each of the symptoms of interest. The case number (1 through 16) corresponds to the time/diagnosis interaction. You will need to compute new variables that separate out the Time and Diagnosis values.

  1. From the menus choose:

    Transform > Compute Variable...

    Figure 3. Compute Variable dialog
    Compute Variable dialog
  2. In the Compute Variable dialog, type time as the target variable.
  3. Type trunc(($casenum-1)/4) + 1 as the numeric expression.
  4. Click OK.
    Figure 4. Compute Variable dialog
    Compute Variable dialog
  5. Recall the Compute Variable dialog.
  6. Type diagnosis as the target variable.
  7. Type mod($casenum-1, 4) + 1 as the numeric expression.
  8. Click OK.
    Figure 5. Projected_centroids.sav
    Variable view of the Data Editor, showing a drop-down list of measurement level choices in the Measure column for the variable diagnosis
  9. In the Variable View for projected_centroids.sav, change the measure for diagnosis from Scale to Nominal.
    Figure 6. Chart Builder
    Chart Builder
  10. Finally, to view the projected centroids of time of diagnosis on binging over time, recall the Chart Builder and click Reset to clear your previous selections.
  11. Select the Scatter/Dot gallery and choose Grouped Scatter.
  12. Select Centroids Projected on Binge eating as the y-axis variable and time as the x-axis variable.
  13. Choose to set colors by diagnosis.
  14. Click OK.
    Figure 7. Projected centroids of Time of diagnosis on Binge eating over time
    Multiple line chart with a separate line for each diagnosis
  15. Then, to connect the points, double-click on the graph, and click the Add interpolation line tool in the Chart Editor.
  16. Close the Chart Editor.

    With respect to binge eating, it is clear that the anorectic groups have different starting values from the bulimic groups. This difference shrinks over time, as the anorectic groups hardly change, while the bulimic groups show progress.

  17. Recall the Chart Builder.
  18. Deselect Centroids Projected on Binge eating as the y-axis variable and select Centroids Projected on Sexual attitude as the y-axis variable.
  19. Click OK.
    Figure 8. Projected centroids of Time of diagnosis on Sexual attitude over time
    Multiple line chart with a separate line for each diagnosis
  20. Then, to connect the points, double-click on the graph, and click the Add interpolation line tool in the Chart Editor.
  21. Close the Chart Editor.

    With respect to sexual attitude, the four trajectories are more or less parallel over time, and all groups show progress. The bulimic groups, however, have higher (better) scores than the anorectic group.

  22. Recall the Chart Builder.
  23. Deselect Centroids Projected on Sexual attitude as the y-axis variable and select Centroids Projected on Preoccupation with food and weight as the y-axis variable.
  24. Click OK.
    Figure 9. Projected centroids of Time of diagnosis on Body preoccupation over time
    Multiple line chart with a separate line for each diagnosis
  25. Then, to connect the points, double-click on the graph, and click the Add interpolation line tool in the Chart Editor.
  26. Close the Chart Editor.

Body preoccupation is a variable that represents the core symptoms, which are shared by the four different groups. Apart from the atypical eating disorder patients, the anorectic group and the two bulimic groups have very similar levels both at the beginning and at the end.