predictBSTFA

Estimate/predict values of the time series at a specific location.

Implements Bayesian spatio-temporal factor analysis models for multivariate data observed across space and time. The package provides tools for model fitting via Markov chain Monte Carlo (MCMC), spatial and temporal interpolation, and visualization of latent factors and loadings to support inference and exploration of underlying spatio-temporal patterns. Designed for use in environmental, ecological, or public health applications, with support for posterior prediction and uncertainty quantification. Includes functions such as BSTFA() for model fitting and plot_factor() to visualize the latent processes. Functions are based on and extended from methods described in Berrett, et al. (2020) <doi:10.1002/env.2609>.

Candace Berrett

BSTFA

Bayesian Spatio-Temporal Factor Analysis Model

Adam Simpson

predictBSTFA function

<dl><dt>out</dt>
<dd>Output from BSTFA or BSTFAfull.</dd>
<dt>location</dt>
<dd>Either a single integer indicating the location in the data set to provide predictions or a vector of length 2 providing the longitude and latitude of the new location. If <code>location=NULL</code> (default), the function will return predictions for all in-sample locations.</dd>
<dt>type</dt>
<dd>One of <code>all</code>, <code>mean</code> (default), <code>median</code>, <code>ub</code>, or <code>lb</code> indicating which summary statistic of the predicted values to return.</dd>
<dt>ci.level</dt>
<dd>If <code>type='lb'</code> or <code>'ub'</code>, the percentiles for the posterior interval.</dd>
<dt>new_x</dt>
<dd>If the original model included covariates <code>x</code>, include the same covariates for prediction <code>location</code>.</dd>
<dt>pred.int</dt>
<dd>Logical scalar indicating whether to include additional uncertainty for posterior predictive intervals (<code>TRUE</code>) or not (<code>FALSE</code>; default -- intervals represent posterior probabilities around the mean of <code>location</code>).</dd></dl>

Arguments

Author

Estimate/predict values of the time series at a specific location. — predictBSTFA

<dl>

<dt>out</dt>
<dd>Output from BSTFA or BSTFAfull.</dd>


<dt>location</dt>
<dd>Either a single integer indicating the location in the data set to provide predictions or a vector of length 2 providing the longitude and latitude of the new location. If <code>location=NULL</code> (default), the function will return predictions for all in-sample locations.</dd>


<dt>type</dt>
<dd>One of <code>all</code>, <code>mean</code> (default), <code>median</code>, <code>ub</code>, or <code>lb</code> indicating which summary statistic of the predicted values to return.</dd>


<dt>ci.level</dt>
<dd>If <code>type='lb'</code> or <code>'ub'</code>, the percentiles for the posterior interval.</dd>


<dt>new_x</dt>
<dd>If the original model included covariates <code>x</code>, include the same covariates for prediction <code>location</code>.</dd>


<dt>pred.int</dt>
<dd>Logical scalar indicating whether to include additional uncertainty for posterior predictive intervals (<code>TRUE</code>) or not (<code>FALSE</code>; default -- intervals represent posterior probabilities around the mean of <code>location</code>).</dd>

</dl>

predictBSTFA: Estimate/predict values of the time series at a specific location.

Description

Usage

Value

Arguments

Author

Examples