2021
DOI: 10.48550/arxiv.2108.02151
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

Semiparametric Functional Factor Models with Bayesian Rank Selection

Abstract: Functional data are frequently accompanied by parametric templates that describe the typical shapes of the functions. Although the templates incorporate critical domain knowledge, parametric functional data models can incur significant bias, which undermines the usefulness and interpretability of these models. To correct for model misspecification, we augment the parametric templates with an infinite-dimensional nonparametric functional basis. Crucially, the nonparametric factors are regularized with an ordere… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1

Citation Types

0
2
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
1
1

Relationship

0
2

Authors

Journals

citations
Cited by 2 publications
(2 citation statements)
references
References 39 publications
0
2
0
Order By: Relevance
“…In fact, this kind of prior is frequently used in a variety of factor models, particularly for high-dimensional inference. Some examples of these shrinkage priors are the Dirichlet-Laplace prior of Bhattacharya et al (2015) or the ordered spike-and-slab prior of Kowal (2021). Unfortunately, these priors do not assist with the inference for Λ in our model.…”
Section: B2 Inference On the Factor Loading Matrixmentioning
confidence: 99%
“…In fact, this kind of prior is frequently used in a variety of factor models, particularly for high-dimensional inference. Some examples of these shrinkage priors are the Dirichlet-Laplace prior of Bhattacharya et al (2015) or the ordered spike-and-slab prior of Kowal (2021). Unfortunately, these priors do not assist with the inference for Λ in our model.…”
Section: B2 Inference On the Factor Loading Matrixmentioning
confidence: 99%
“…This can be achieved by regarding f i (τ j ) = λ m (τ j )η im as a linear combination of (orthogonal) basis functions, often modelled using splines, which then implies a particular covariance function for f i (τ ). Although some of the research in this direction uses a Bayesian approach to inference (Kowal et al 2017;Kowal and Canale 2022), the frequentist treatment of the problem is more common, particularly in financial applications to forecasting yield curves (Hays et al 2012;Jungbacker et al 2014). An alternative approach in models with functional factor loadings is to assume each λ m (τ ) to be a realisation of a stochastic process, for instance, a Gaussian process with a chosen covariance function; see, for example, the dynamic spatial model of Lopes et al (2008).…”
Section: Introductionmentioning
confidence: 99%