Inverse prediction (IP) is a statistical methodology that can be used to obtain confidence sets on estimates. This project's objective was to extend and adapt methods of IP for use in forensically important settings. The three main parts of this research were to develop methods of IP multivariate quantitative responses with heterogeneous variance-covariance structure, including flexible, adaptive modeling of both means and variances; to develop methods of IP for categorical responses with the same modeling flexibility; and to integrate these results to create IP methods for hybrid quantitative-categorical responses. A common objective in this three-pronged effort was to assess and quantify the inherent uncertainty accurately and defensibly. The over-arching objective was to develop a set of statistical methods and versatile, accessible tools for IP in settings important in assessing time since death and other forensically important applications. This objective was accomplished by making it possible for IP methodology to be implemented within the broad context of mixed linear models. Its computations leading to p-values and confidence sets can be more easily performed, because an investigator can now do the analysis using a variety of widely available statistical computing packages. The project has submitted for publication scientific manuscripts to illustrate the immediate practical value of the IP methods for guiding PMI research design. Scholarly products are listed.
Downloads
Similar Publications
- Just Science Podcast: Just Rapidly Identifying Drugs Involved in Suspected Overdoses
- Decreased Accuracy of Forensic DNA Mixture Analysis for Groups with Lower Genetic Diversity
- Extinction Training Suppresses Activity of Fear Memory Ensembles across the Hippocampus and Alters Transcriptomes of Fear-encoding Cells