Reconstructability Analysis and Log-Linear Modeling

Authors

Martin Zwick, Portland State UniversityFollow

Document Type

Presentation

Publication Date

4-16-2005

Subjects

System design, System analysis, Log-linear models, Data structures (Computer science)

Abstract

Reconstructability Analysis (RA) is a method developed within the systems community for analyzing nominal (or discretized) data. RA both overlaps and extends Log-Linear (LL) modeling, and the purpose of this talk is to introduce RA to researchers unfamiliar with it. Two aspects of RA will be focused on: (1) its use for exploratory, as opposed to confirmatory, modeling – searching for good models in a vast space of possible models, and (2) state-based RA – analyzing data not in terms of relations among variables but in terms of relations among specific states of variables. Examples of applications to social science data and software implementations will be discussed. Other features of RA, e.g., its use of the intuitively graspable uncertainty reduction measure, its application to the merging of multiple datasets whose variables overlap, and its non-statistical forms, will also be mentioned.

Keywords: Reconstructability Analysis, Log-Linear Modeling, categorical data, multivariate statistics, exploratory modeling, confirmatory modeling

Rights

© The Author

Description

Talk given to Western Psychology Association meeting, 4/16/2005

Persistent Identifier

https://archives.pdx.edu/ds/psu/42795

Citation Details

Martin Zwick (2005). "Reconstructability Analysis and Log-Linear Modeling." Western Psychological Association, Portland, Oregon, 4/16/2005.