Mark Stemmler – författare

This book offers a comprehensible overview of the statistical approach called the person-centered method. Instead of analyzing means, variances and covariances of scale scores as in the common variable-centered approach, the person-centered approach analyzes persons or objects grouped according to their characteristic patterns or configurations in contingency tables. This second edition explores the relationship between two statistical methods: log-linear modeling (LLM) and configural frequency analysis (CFA). Both methods compare expected frequencies with observed frequencies. However, while LLM searches for the underlying dependencies of the involved variables in the data (model-fitting), CFA examines significant residuals in non-fitting models.

New developments in the second edition include: Configural Mediation Models, CFA with covariates, moderator CFA, and CFA modeling branches in tree-based methods. The new developments enable the use of categorical together with continuous variables, which makes CFA a very powerful statistical tool. This new edition continues to utilize R-package confreq (derived from Configural Frequency Analysis), much updated since the first edition and newly adjusted to the new R base program 4.0. An electronic supplement is now available with 18 R-scripts and many datasets.

This book covers the following subjects: growth curve modeling, directional dependence, dyadic data modeling, item response modeling (IRT), and other methods for the analysis of dependent data (e.g., approaches for modeling cross-section dependence, multidimensional scaling techniques, and mixed models). It presents contributions on handling data in which the postulate of independence in the data matrix is violated. When this postulate is violated and when the methods assuming independence are still applied, the estimated parameters are likely to be biased, and statistical decisions are very likely to be incorrect. Problems associated with dependence in data have been known for a long time, and led to the development of tailored methods for the analysis of dependent data in various areas of statistical analysis. These include, for example, methods for the analysis of longitudinal data, corrections for dependency, and corrections for degrees of freedom. 

Researchers and graduate students in the social and behavioral sciences, education, econometrics, and medicine will find this up-to-date overview of modern statistical approaches for dealing with problems related to dependent data particularly useful.

This book takes an easy-to-understand look at the statistical approach called the person-centered method. Instead of analyzing means, variances and covariances of scale scores as in the common variable-centered approach, the person-centered approach analyzes persons or objects grouped according to their characteristic patterns or configurations in contingency tables. The main focus of the book will be on Configural Frequency Analysis (CFA; Lienert and Krauth, 1975) which is a statistical method that looks for over and under-frequented cells or patterns. Over frequented means that the observations in this cell or configuration are observed more often than expected, under-frequented means that this cell or configuration is observed less often than expected. In CFA a pattern or configuration that contains more observed cases than expected is called a type; similarly, a pattern or configuration that is less observed than expected are called an antitype. CFA is similar to log-linear modeling. In log-linear modeling the goal is to come up with a fitting model including all important variables. Instead of fitting a model, CFA looks at the significant residuals of a log-linear model.

The book describes the use of an R-package called confreq (derived from Configural Frequency Analysis). The use of the software package is described and demonstrated with data examples.

This volume presents contributions on handling data in which the postulate of independence in the data matrix is violated. When this postulate is violated and when the methods assuming independence are still applied, the estimated parameters are likely to be biased, and statistical decisions are very likely to be incorrect. Problems associated with dependence in data have been known for a long time, and led to the development of tailored methods for the analysis of dependent data in various areas of statistical analysis. These methods include, for example, methods for the analysis of longitudinal data, corrections for dependency, and corrections for degrees of freedom. This volume contains the following five sections: growth curve modeling, directional dependence, dyadic data modeling, item response modeling (IRT), and other methods for the analysis of dependent data (e.g., approaches for modeling cross-section dependence, multidimensional scaling techniques, and mixed models). Researchers and graduate students in the social and behavioral sciences, education, econometrics, and medicine will find this up-to-date overview of modern statistical approaches for dealing with problems related to dependent data particularly useful.