Regression analysis of multiple outcomes Modelli di regressione per risposte multiple (original) (raw)
2015
Abstract. The hierarchical linear model in a linear model with nested random coefficients, fruitfully used for multilevel research. A tutorial is presented on the use of this model for the analysis of longitudinal data, i.e., repeated ata on the same subjects. An important advantage of this approach is that differences across subjects in the numbers and spacings of measurement occasions do not present a problem, and that changing covariates can easily be handled. The tutorial approaches the longitudinal data as measurements on populations of (subject-specific) functions. Key words. Multilevel analysis, hierarchical linear model, random coefficients. 1. When and why use the hierarchical inear model for analyzing longitudinal data? A large variety of statistical methods exists for the analysis of longitudinal data. This paper is a tutorial that explains the use of the hierarchical linear model, also referred to as the multilevel model, for analysing longitudinal data. The hierarchical...
This series of articles, which provide an overview of several advanced statistical methods for evaluating treatment outcomes and mechanisms of change, makes up the first research methods– oriented special issue to appear in the Journal of Consulting and Clinical Psychology. Like most active areas of scientific inquiry, the field of biostatistics and quantitative psychology continues to innovate and progress at a remarkable pace. These recent advancements provide researchers with the tools needed to ask and get answers to progressively nuanced and complex questions. It is our hope that the articles included in this special issue will spark an interest among applied researchers to investigate and learn how these and other newer statistical methods might be used to enhance their own line of research. This special issue, " Advances in Data Analytic Methods for Evaluating Treatment Outcome and Mechanisms of Change, " is a collection of articles illustrating the application of a broad range of cutting-edge statistical methods. The field of applied biostatistics has been evolving at a rapid pace, and keeping current with these advancements is a challenge for all psychologists. Our primary goal for this special issue is to provide the readership of the Journal of Consulting and Clinical Psychology with a sample of several recent developments within biostatistics, developments that are likely to be especially relevant to applied clinical researchers. The call for papers for this special issue resulted in more than 120 abstract submissions from a wide variety of clinical and quantitative researchers. From this pool, 20 abstracts were selected, and the authors were invited to submit full papers for peer review. Fourteen papers successfully made it through the peer review process and are presented in this special issue. Most of the selected articles include an illustration of the statistical method using real data (or appropriate simulation studies) relevant to clinical researchers and, when appropriate, code is provided for common statistical software packages, a feature that will be helpful for readers interested in applying these methods to their own data. The articles in this special issue include novel approaches to modeling multiple outcomes, exploring predictors of treatment outcome, testing causality, assessing potential bias resulting from missing data, and designing adaptive interventions using methods adapted from control engineering. Wright, Hallquist, Swartz, Frank, and Cyranowski (2014, this issue) and Shiyko, Burkhalter, Li, and Park (2014, this issue) illustrate how time-varying effect models (TVEMs) can be used to understand the effect of treatment as it changes dynamically over time. Wright et al. showed how such models can be used to explore the dynamic relationship between depression and anxiety within the context of a psychotherapy trial. Shiyko et al. used data from a study of lung cancer patients recovering from two different types of surgical procedures and demonstrated how TVEMs can be used to evaluate whether the effect of treatment on outcomes changes over time. The authors also showed how TVEMs can address many of the statistical challenges inherent in time-intensive longitudinal data, such as large number of assessments , incomplete and unbalanced data, and nonlinear response patterns. Ramseyer, Kupper, Caspar, Znoj, and Tschacher (2014, this issue) illustrated the use of time series panel analysis , which can be used to investigate the temporal interplay between multiple variables in time-intensive longitudinal data, such as long-term psychotherapy and diary studies. Finally, Lagoa, Bekiroglu, Lanza, and Murphy (2014, this issue) showed how methods borrowed from control engineering can be used with intensive longitudinal data to design and evaluate adaptive interventions (or sequences of treatments) that are modified to individual response and circumstances. Missing data have long presented an analytic challenge. The majority of statistical methods used by clinical researchers assume that missing data are missing at random (MAR), an untestable assumption.