Mr Víctor Miranda

R, C++, SAS, Wolfram Mathematica, Linux (Fedora, Red Hat)

Profile Image
Doctoral Candidate - Doctor of Philosophy


Pursuing PhD, Victor joined the UoA Department of Statistics in September 2014 under supervision of Dr Thomas W. Yee. Currently allocated to Room 303s-384, Victor has a strong background on Maths and Statistics. While working in Merida, Mexico, as lecturer (2005) at the University of Yucatan (UoY), he obtained the Bachelor Degree on Mathematics (UoY), and then a MSc on Statistics (Major: Computational Statistics) from the National Autonomous University of Mexico (2008).  From 2008 -- 011, he was appointed as Biostatistician by the National Institute of Public Health (Cuernavaca, Mexico) to research on topics related to Breast Cancer, Nutrition, Obesity, and Air Pollution.  Generalized Additive Models, Mixed Models, Time Series Analysis using Distributed Lag Models (Zanobetti & Schwartz, 2000), and Meta-Regression Analyses were included on his research responsibilities. From 2011 to August 2014 he worked for the University of Quintana Roo, Department of Mathematics, with special interest on time series modelling frameworks, vector Generalized Additive models, generalized logistic estimation on finite populations, and Numerical analysis on ecological models. Estimated time for PhD thesis submission: April 2018.

Research | Current

Victor's work concentrates on extending the VGAM 'R' package, from Thomas W. Yee, by implementing a new class of VGLMs/ VGAMs to handle time series (TS) models via object oriented methods in S4. Unlike other software to fit this class of data (particularly in R, those relying on optim() and arima()), both, the theory and the computational structure rely on MLE using Fisher scoring. This framework is designed to fit popular TS models (linear and non-linear, e.g., IGARCHs) as special cases, and accommodates covariates in the analysis allowing to model empirical features of stochastic volatility, unrivalled by other methodology. In addition, his work aims to show how VGLMs and Reduced-Ranks VGLMs (Yee & Hastie, 2004) straightforwardly manage cointegrated time series, with little literature addressing such topic (e.g., Pfaff, 2011). Finally, Victor is developing new VGLM link functions for the mean and the quantiles of several 1-parameter distributions. The latter arises as an alternative to quantile regression, exceedingly addressed in the literature with many variants, but no overriding framework.

Teaching | Current

Graduate Teaching Assitant.

Postgraduate supervision

Dr Thomas Yee and Dr Yong Wang