Loading…
This event has ended. Visit the official site or create your own event on Sched.
Click here to return to main conference site. For a one page, printable overview of the schedule, see this.
View analytic
Wednesday, June 29 • 2:30pm - 3:30pm
hurdlr: An R package for zero-inflated and over-dispersed count data

Log in to save this to your schedule and see who's attending!

Poster #10

When considering count data, it is often the case that many more zero counts than would be expected of some given distribution are observed. It is well-established that data such as this can be reliably modeled using zero-inflated or hurdle distributions. However, it is also not uncommon that count data, especially ecological or environmental data, contain some number of extremely large observations which typically would be considered outliers and excluded from analyses due to difficulties in model fitting.In lieu of throwing out data or risk mis-specifying the distributional form, observations above a given threshold can be modeled by, e.g., an extreme value distribution, or any distribution with a long right tail. To utilize this modeling technique, we develop an R package ``hurdlr'' to utilize the double-hurdle model of Balderama, Gardner, and Reich (2014), which accounts for both the zero-inflation and extreme over-dispersion present in many count data sets. The hurdlr package functions are flexible and versatile: it can be applied with various count distributions and are able to allow for one or multiple hurdles. A Bayesian hierarchical framework is used for estimation, and covariate information can be included to inform top-level parameters.

Speakers
EB

Earvin Balderama

Loyola University Chicago, Chicago, IL


Wednesday June 29, 2016 2:30pm - 3:30pm
Sponsor Pavilion 326 Galvez Street Stanford, CA 94305-6105

Attendees (31)