A Projection Approach to Local Regression with Variable-Dimension Covariates

Simple item page
dc.contributor.authorHeiner, Matthew J.
dc.contributor.authorPage, Garritt L.
dc.contributor.authorQuintana, Fernando Andres
dc.date.accessioned2025-01-20T16:15:34Z
dc.date.available2025-01-20T16:15:34Z
dc.date.issued2024
dc.description.abstractIncomplete covariate vectors are known to be problematic for estimation and inferences on model parameters, but their impact on prediction performance is less understood. We develop an imputation-free method that builds on a random partition model admitting variable-dimension covariates. Cluster-specific response models further incorporate covariates via linear predictors, facilitating estimation of smooth prediction surfaces with relatively few clusters. We exploit marginalization techniques of Gaussian kernels to analytically project response distributions according to any pattern of missing covariates, yielding a local regression with internally consistent uncertainty propagation that uses only one set of coefficients per cluster. Aggressive shrinkage of these coefficients regulates uncertainty due to missing covariates. The method allows in- and out-of-sample prediction for any missingness pattern, even if the pattern in a new subject's incomplete covariate vector was not seen in the training data. We develop an MCMC algorithm for posterior sampling that improves a computationally expensive update for latent cluster allocation. Finally, we demonstrate the model's effectiveness for nonlinear point and density prediction under various circumstances by comparing with other recent methods for regression of variable dimensions on synthetic and real data. Supplemental materials for this article are available online.
dc.description.funderGrant FONDECYT
dc.fuente.origenWOS
dc.identifier.doi10.1080/10618600.2024.2357636
dc.identifier.eissn1537-2715
dc.identifier.issn1061-8600
dc.identifier.urihttps://doi.org/10.1080/10618600.2024.2357636
dc.identifier.urihttps://repositorio.uc.cl/handle/11534/90501
dc.identifier.wosidWOS:001249015000001
dc.language.isoen
dc.revistaJournal of computational and graphical statistics
dc.rightsacceso restringido
dc.subjectBayesian nonparametrics
dc.subjectClustering
dc.subjectDependent random partition models
dc.subjectIndicator missing
dc.subjectPattern missing
dc.subject.ods03 Good Health and Well-being
dc.subject.odspa03 Salud y bienestar
dc.titleA Projection Approach to Local Regression with Variable-Dimension Covariates
dc.typeartículo
sipa.indexWOS
sipa.trazabilidadWOS;2025-01-12
Files
Collections
Artículos de revistas