Author : Erkan Sayilir
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (134 download)
Book Synopsis Bayesian Modeling of Single Case Research Design Data in the Presence of Autocorrelation by : Erkan Sayilir
Download or read book Bayesian Modeling of Single Case Research Design Data in the Presence of Autocorrelation written by Erkan Sayilir and published by . This book was released on 2020 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: In the current study, I intended to simulate single case research design (SCRD) data to investigate the impact of the presence of autocorrelation on analysis of SCRD for Bayesian method under a variety of simulation conditions. The purpose of this study is to analyze the results of data obtained in the context of an individual single-case study, for a variety of simulation conditions such as varying sample sizes, varying numbers of measurement observations, and the presence of varied levels of autocorrelation. Then I compared performance of both methods using Bias, Percentage Bias (PB), Mean Square Error (MSE), Power, and DIC statistics. Analysis of variance (ANOVA) was implemented to analyze the summary statistics and compare performance of both approaches. Another goal of this study was to develop and fit several Bayesian models from simple model to a complex model to evaluate the significance of the model parameters in a SCRD in the presence of autocorrelation. Also, the effect of using a scaled-inverse Wishart distribution over parameter estimation as a prior distribution for covariance matrix of parameters was explored. It was found that the performance of the Bayesian models in this study was satisfactory, except that of the Scaled inverse-Wishart distribution with autocorrelated error (Model 3B). It produced extremely large biased estimates, percentage bias (PB), standardized bias (SB) and mean square error (MSE) for most of the simulation conditions. On the other hand, the Scaled-inverse Wishart distribution performed well if serial dependency was not taken into account across simulation conditions. Also, the SB was around 5%, bias still had an adverse effect on the coverage rates of the models. A poor convergence rate was observed for SIW regardless of modeling autocorrelation and convergence was not reached for Random Slopes Model (RSM) for some of the simulation conditions when autocorrelation was considered. Although mean standardized biases of the models were around 5%, it is likely that having larger standard error for the models had effects on the power and the coverage rates of the models. In most cases, coverage rate under 95% were observed, which means under-coverage and indicates higher Type- I error rate should be expected.