What we’re about
Our purpose is to build community, share knowledge, and grow influence for users and producers of federal statistics.
Anyone interested in open source tools, R programming and federal statistics is welcome to participate.
We will host workshops, seminars, and social gatherings. We will also curate resources.
Membership is open to all and there is no membership fee.
R Govys is an R User Group sponsored by the R Consortium.
Upcoming events (1)
See all- R Govys: Fellegi-Sunter Decisions beyond Pairwise Record LinkageLink visible for attendees
Fellegi and Sunter (1969) formalize the probabilistic theory for linking pairs of records. Their theory gave rise to a powerful methodology based on expanding exhaustive classifications of record pairs. This approach is powerful and widely used but has implicit limitations. Sadinle and Fienberg (2013) expose explicitly the limitations of linking general configurations of records proceeding strictly from classified pairs. They propose an expanded universe of linkage structures transcending the limitations of pair-based linkage. I discuss simple instances of such linkage structures -record triples. I show how binary Fellegi-Sunter decisions -"match or "nonmatch"- can be applied successively to take advantage of these more complex structures to construct principled sets of linked records. Tools and approaches for record linkage in R will be used to convey the concepts.
Presented by: Yves Thibaudeau, Center for Statistical Research and Methodology, U.S. Census Bureau ~ After completing a PhD at Carnegie Mellon University in 1988, Yves joined the record-linkage group headed by Bill Winkler in the Statistical Research Division. Over his career Yves has worked on problems involving missing data and classification, in addition to record linkage.
Register here: https://amstat.zoom.us/webinar/register/WN_emap57YPSkGWxZ8muurqKA
