• EN
  • ES


Monday, June 24, 2019

Seminario de Investigación INAMAT: Mitigating bias from unobserved spatial confounders using linear mixed models

Por Patrick Schnell

El INAMAT organiza el Seminario de investigación titulado: Mitigating bias from unobserved spatial confounders using linear mixed models impartido por Patrick Schnell de la Universidad de Ohio.

Fecha: 24 de Junio a las 10 am
Lugar: Sala de Conferencias Jerónimo de Ayanz


Spatially correlated random effects are often added to models for spatially referenced data to account for unobserved variables that are also assumed to be spatially correlated. In standard models, these spatial random effects are independent of the covariates a priori. If the unobserved variables are confounders, i.e., they are correlated with both the covariates and the outcome, the standard usage of spatial random effects fails to mitigate the resulting bias. We present an approach for mitigating bias due to unobserved spatial confounders by linking the spatial random effects to covariates in a joint model, and give examples of assumptions under which the parameters necessary for bias mitigation are identifiable. Our approach is illustrated in the study of the effects of a neighborhood’s median household income on crime in Columbus, Ohio, United States

Additional documentation:

Research Institutes
Jeronimo de Ayanz Building
Public University of Navarre
Campus de Arrosadia 31006 - Pamplona
Tel. +34 948 169512
Contact by E-mail