This Research Topic in Frontiers in Quantitative Psychology and Measurement will focus on the methodological issues pertaining specifically to statistical forms of classification and showcase newer advancements in statistical classification analysis.  This issue will feature discussion pieces about the nature of and issues associated with classification as well as both simulation studies of statistical properties of classification analyses and papers presenting applications of new and cutting edge forms of statistical classification.   The purpose for this issue is to present a collection of papers which provide the reader with a holistic view of the current state and potential future of classification analysis.

Ball State University

Frontiers in Psychology

Quantitative Psychology and Measurement

Frontiers in Applied Mathematics and Statistics

Methodological Issues and Advancements in Classification

Background: Informant discrepancies have been reported between parent and adolescent measures of depressive disorders and suicidality. We aimed to examine the concordance between adolescent and parent ratings of depressive disorder using both clinical interview and questionnaire measures and assess multi-informant and multi-method approaches to classification.Method: Within the context of assessment of eligibility for a randomized clinical trial, 50 parent–adolescent pairs (mean age of adolescents = 15.0 years) were interviewed separately with a structured diagnostic interview for depression, the KID-SCID. Adolescent self-report and parent-report versions of the Strengths and Difficulties Questionnaire, the Short Mood and Feelings Questionnaire and the Depressive Experiences Questionnaire were also administered. We examined the diagnostic concordance rates of the parent vs. adolescent structured interview methods and the prediction of adolescent diagnosis via questionnaire methods.Results: Parent proxy reporting of adolescent depression and suicidal thoughts and behavior is not strongly concordant with adolescent report. Adolescent self-reported symptoms on depression scales provide a more accurate report of diagnosable adolescent depression than parent proxy reports of adolescent depressive symptoms. Adolescent self-report measures can be combined to improve the accuracy of classification. Parents tend to over report their adolescent’s depressive symptoms while under reporting their suicidal thoughts and behavior.Conclusion: Parent proxy report is clearly less reliable than the adolescent’s own report of their symptoms and subjective experiences, and could be considered inaccurate for research purposes. While parent report would still be sought clinically where an adolescent refuses to provide information, our findings suggest that parent reporting of adolescent suicidality should be interpreted with caution.

Background: Informant discrepancies have been reported between parent and adolescent measures of depressive disorders and suicidality. We aimed to examine the concordance between adolescent and parent ratings of depressive disorder using both clinical interview and questionnaire measures and assess multi-informant and multi-method approaches to classification.Method: Within the context of assessment of eligibility for a randomized clinical trial, 50 parent–adolescent pairs (mean age of adolescents = 15.0 years) were interviewed separately with a structured diagnostic interview for depression, the KID-SCID. Adolescent self-report and parent-report versions of the Strengths and Difficulties Questionnaire, the Short Mood and Feelings Questionnaire and the Depressive Experiences Questionnaire were also administered. We examined the diagnostic concordance rates of the parent vs. adolescent structured interview methods and the prediction of adolescent diagnosis via questionnaire methods.Results: Parent proxy reporting of adolescent depression and suicidal thoughts and behavior is not strongly concordant with adolescent report. Adolescent self-reported symptoms on depression scales provide a more accurate report of diagnosable adolescent depression than parent proxy reports of adolescent depressive symptoms. Adolescent self-report measures can be combined to improve the accuracy of classification. Parents tend to over report their adolescent’s depressive symptoms while under reporting their suicidal thoughts and behavior.Conclusion: Parent proxy report is clearly less reliable than the adolescent’s own report of their symptoms and subjective experiences, and could be considered inaccurate for research purposes. While parent report would still be sought clinically where an adolescent refuses to provide information, our findings suggest that parent reporting of adolescent suicidality should be interpreted with caution.

Depression and suicidal behavior in adolescents: a multi-informant and multi-methods approach to diagnostic classification

Classification using standard statistical methods such as linear discriminant analysis (LDA) or logistic regression (LR) presume knowledge of group membership prior to the development of an algorithm for prediction. However, in many real world applications members of the same nominal group, might in fact come from different subpopulations on the underlying construct. For example, individuals diagnosed with depression will not all have the same levels of this disorder, though for the purposes of LDA or LR they will be treated in the same manner. The goal of this simulation study was to examine the performance of several methods for group classification in the case where within group membership was not homogeneous. For example, suppose there are 3 known groups but within each group two unknown classes. Several approaches were compared, including LDA, LR, classification and regression trees (CART), generalized additive models (GAM), and mixture discriminant analysis (MIXDA). Results of the study indicated that CART and mixture discriminant analysis were the most effective tools for situations in which known groups were not homogeneous, whereas LDA, LR, and GAM had the highest rates of misclassification. Implications of these results for theory and practice are discussed.

Classification using standard statistical methods such as linear discriminant analysis (LDA) or logistic regression (LR) presume knowledge of group membership prior to the development of an algorithm for prediction. However, in many real world applications members of the same nominal group, might in fact come from different subpopulations on the underlying construct. For example, individuals diagnosed with depression will not all have the same levels of this disorder, though for the purposes of LDA or LR they will be treated in the same manner. The goal of this simulation study was to examine the performance of several methods for group classification in the case where within group membership was not homogeneous. For example, suppose there are 3 known groups but within each group two unknown classes. Several approaches were compared, including LDA, LR, classification and regression trees (CART), generalized additive models (GAM), and mixture discriminant analysis (MIXDA). Results of the study indicated that CART and mixture discriminant analysis were the most effective tools for situations in which known groups were not homogeneous, whereas LDA, LR, and GAM had the highest rates of misclassification. Implications of these results for theory and practice are discussed.

Group membership prediction when known groups consist of unknown subgroups: a Monte Carlo comparison of methods

Although traditional clustering methods (e.g., K-means) have been shown to be useful in the social sciences it is often difficult for such methods to handle situations where clusters in the population overlap or are ambiguous. Fuzzy clustering, a method already recognized in many disciplines, provides a more flexible alternative to these traditional clustering methods. Fuzzy clustering differs from other traditional clustering methods in that it allows for a case to belong to multiple clusters simultaneously. Unfortunately, fuzzy clustering techniques remain relatively unused in the social and behavioral sciences. The purpose of this paper is to introduce fuzzy clustering to these audiences who are currently relatively unfamiliar with the technique. In order to demonstrate the advantages associated with this method, cluster solutions of a common perfectionism measure were created using both fuzzy clustering and K-means clustering, and the results compared. Results of these analyses reveal that different cluster solutions are found by the two methods, and the similarity between the different clustering solutions depends on the amount of cluster overlap allowed for in fuzzy clustering.

Although traditional clustering methods (e.g., K-means) have been shown to be useful in the social sciences it is often difficult for such methods to handle situations where clusters in the population overlap or are ambiguous. Fuzzy clustering, a method already recognized in many disciplines, provides a more flexible alternative to these traditional clustering methods. Fuzzy clustering differs from other traditional clustering methods in that it allows for a case to belong to multiple clusters simultaneously. Unfortunately, fuzzy clustering techniques remain relatively unused in the social and behavioral sciences. The purpose of this paper is to introduce fuzzy clustering to these audiences who are currently relatively unfamiliar with the technique. In order to demonstrate the advantages associated with this method, cluster solutions of a common perfectionism measure were created using both fuzzy clustering and K-means clustering, and the results compared. Results of these analyses reveal that different cluster solutions are found by the two methods, and the similarity between the different clustering solutions depends on the amount of cluster overlap allowed for in fuzzy clustering.

Applications of cluster analysis to the creation of perfectionism profiles: a comparison of two clustering approaches

Statistical classification of phenomena into observed groups is very common in the social and behavioral sciences. Statistical classification methods, however, are affected by the characteristics of the data under study. Statistical classification can be further complicated by initial misclassification of the observed groups. The purpose of this study is to investigate the impact of initial training data misclassification on several statistical classification and data mining techniques. Misclassification conditions in the three group case will be simulated and results will be presented in terms of overall as well as subgroup classification accuracy. Results show decreased classification accuracy as sample size, group separation and group size ratio decrease and as misclassification percentage increases with random forests demonstrating the highest accuracy across conditions.

Statistical classification of phenomena into observed groups is very common in the social and behavioral sciences. Statistical classification methods, however, are affected by the characteristics of the data under study. Statistical classification can be further complicated by initial misclassification of the observed groups. The purpose of this study is to investigate the impact of initial training data misclassification on several statistical classification and data mining techniques. Misclassification conditions in the three group case will be simulated and results will be presented in terms of overall as well as subgroup classification accuracy. Results show decreased classification accuracy as sample size, group separation and group size ratio decrease and as misclassification percentage increases with random forests demonstrating the highest accuracy across conditions.