This project will develop and implement several novel approaches of finite mixture modeling to psychological diagnosis and educational assessment. In extending this methodology to psychological applications, several modifications of the finite mixture models will be made to address the differing data types, model demands, and diagnostic characteristics (for instance, as provided by the Diagnostic and Statistical Manual of Mental Disorders) present in psychological research. Specific objectives of the research are: (1) develop constrained finite mixture models for psychological measurement and diagnosis, (2) develop the capability of utilizing multiple data types to more efficiently determine profiles underlying the criteria specified in the Diagnostic and Statistical Manual of Mental Disorders, (3) construct exploratory procedures for linking behavioral data to diagnostic criteria defined by the Diagnostic and Statistical Manual of Mental Disorders, and (4) provide widely available software for practitioners to utilize these new models.
The objectives of this research will result in contributions to psychology, educational assessment, and statistics. In psychology, the Diagnostic and Statistical Manual of Mental Disorders defines pathology based on a profile of criteria that have or have not been met. Typically, there are multiple profiles that can lead to the same diagnosis (pathological or not), where patients can benefit from being differentiated based on their profile to better select treatment options. This research will develop understanding of the complex nature of psychological disorders and the resulting treatment needs of the differing types of patients. Such methods will be beneficial to many other areas of psychology, such as assessment of developmental changes and the study of personality traits. In education, these methods will further the ability to provide students with diagnostic feedback about their performance, providing avenues for student improvement and remediation. These tools extend beyond traditional testing formats (e.g. paper and pencil tests) and typical examinee populations, allowing for alternative methods of assessment. In statistics, this project broadens a set of procedures used for classification. Software developed as part of the research will enable researchers and practitioners to implement such methods in a wide spectrum of applications.
