| One of the debates about personality disorders has been whether they are categorical or dimensional constructs. This study used an empirical method, Maximum Covariance Analysis (MAXCOV), to attempt to determine whether Borderline Personality Disorder (BPD) matches a categorical or dimensional model. MAXCOV was applied to eight variables representing diagnostic criteria for BPD, derived from the Personality Diagnostic Questionnaire (PDQ), a self-report inventory of personality disorders. Data used were previously collected from 552 patients for a validation study of the PDQ.;MAXCOV results were analyzed by a standard method: smoothing the data points and plotting the results. The shape of the resulting curve (peaked or flat) would indicate a categorical or dimensional model, respectively. For the BPD data, the plot was nearly a straight line, consistent with a dimensional model. Because MAXCOV analysis using smoothing and visual inspection does not allow calculation of error probabilities, the accuracy of the BPD analysis is unknown. Consequently, MAXCOV was performed on simulated data that corresponded to dimensional and categorical cases, including data matching the base rates and correlations of the real BPD data. It was discovered that smoothing and visual inspection of a graph can lead to erroneous conclusions. To correct this problem, a new method of interpreting MAXCOV results was developed. Discriminant analysis was applied to the MAXCOV numbers from simulated data, and a discriminant function was found which distinguished the categorical from dimensional cases. The BPD data were then classified using this function. Results of this analysis provided additional evidence for the dimensional model of BPD.;These findings support the recommendation that dimensional constructs be included in future diagnostic systems, in contrast to current systems, which use a purely categorical model of personality disorders. Furthermore, the simulation results raise questions about the validity of the MAXCOV method using smoothing and visual inspection. |
