Journals Information
Mathematics and Statistics Vol. 10(3), pp. 562 - 574
DOI: 10.13189/ms.2022.100312
Reprint (PDF) (1202Kb)
Traumatic Systolic Blood Pressure Modeling: A Spectral Gaussian Process Regression Approach with Robust Sample Covariates
David Kwamena Mensah *, Michael Arthur Ofori , Nathaniel Howard
Department of Statistics, University of Cape Coast, Cape Coast, Ghana
ABSTRACT
Physiological vital signs acquired during traumatic events are informative on the dynamics of the trauma and their relationship with other features such as sample-specific covariates. Non-time dependent covariates may introduce extra challenges in the Gaussian Process () regression, as their main predictors are functions of time. In this regard, the paper introduces the use of Orthogonalized Gnanadesikan-Kettering covariates for handling such predictors within the Gaussian process regression framework. Spectral Bayesian
regression is usually based on symmetric spectral frequencies and this may be too restrictive in some applications, especially physiological vital signs modeling. This paper builds on a fast non-standard variational Bayes method using a modified Van der Waerden sparse spectral approximation that allows uncertainty in covariance function hyperparameters to be handled in a standard way. This allows easy extension of Bayesian methods to complex models where non-time dependent predictors are available and the relationship between the smoothness of trend and covariates is of interest. The utility of the methods is illustrated using both simulations and real traumatic systolic blood pressure time series data.
KEYWORDS
Gaussian Process Regression, Variational Bayes, Traumatic Events, Orthogonalized Gnanadesikan-Kettering Covariates, Systolic Blood Pressure, Smoothness Parameter
Cite This Paper in IEEE or APA Citation Styles
(a). IEEE Format:
[1] David Kwamena Mensah , Michael Arthur Ofori , Nathaniel Howard , "Traumatic Systolic Blood Pressure Modeling: A Spectral Gaussian Process Regression Approach with Robust Sample Covariates," Mathematics and Statistics, Vol. 10, No. 3, pp. 562 - 574, 2022. DOI: 10.13189/ms.2022.100312.
(b). APA Format:
David Kwamena Mensah , Michael Arthur Ofori , Nathaniel Howard (2022). Traumatic Systolic Blood Pressure Modeling: A Spectral Gaussian Process Regression Approach with Robust Sample Covariates. Mathematics and Statistics, 10(3), 562 - 574. DOI: 10.13189/ms.2022.100312.