Teaching
NIMBLE tutorials
All the code and datasets used for the tutorial can be also found here.
In this tutorial I use the famours Scottish lip cancer dataset and provide 4 different approaches for disease mapping. I start with a basic global smoothing by incorporating unstructured random effect, then I show how to fit an ICAR, I continue showwing a BYM model and finally I show how to fit the BYM2 in NIMBLE. The results of the Scottish lip cancer data are then compared and discussed.
This tutorial is motivated by an analysis I conducted to examine the effect of long-term exposure to air-pollution on COVID-19 mortality. The purpose of this tutorial is to show how to use NIMBLE to perform ecological regression and also interpret the results. I am using the BYM2 prior to capture spatial autocorrelation.
This tutorial compares the Leroux model using CARBayes and NIMBLE. It uses the Scottish lip cancer as a case study, shows the code for both approaches and makes relevant comparison for the random fields and hyperparameters across the different softwares.
In this tutorial, I am showing how to fit splines and random walk processes using NIMBLE. In particular, it shows code for cubic splines and random walks of order 1 and 2. The case example is the Chicago dataset to assess the effect of temperature on all cause mortality in Chicago, as used in the vignette of dlnm by Gasparrini.
In this tutorial, I am showing how to fit inseparable spacetime models. I am modeling cervical cancers among HIV positive women in South Africa, as analysed here. I am considering spacetime interaction type I as introduced by Held.
In this tutorial, I am showing how to perform imputation in the Bayesian framework. The case example is the Chicago dataset that was also used on the 4th tutorial and I am modeling the effect of temperature on all cause mortality, while adjusting and imputing for PM$_{10}$.
In this tutorial, I am showing how to fit piecewise linear models in NIMBLE. The threshold is selected in two ways, either by minimizing the WAIC or by adding a random variable for the threshold. I apply the methodology to the Chicago dataset.
INLA tutorials
All the code and datasets used for the tutorial can be also found here.
In this tutorial I use lung cancer mortality in England and I show how to use INLA to perform disease mapping. I fit two different models: the first is with an unstrucutred random effect, whereas the second with the BYM2 prior. I use PC priors as the selected priors and show how to postprocess the INLA output.
In this tutorial I use data on stroke mortality in Shefield together with information about air-pollution levels, and deprivation. I show how to perform ecological regression with INLA and BYM2 and interpret the results accordingly.
In this tutorial I used simulated data to mimic the type 1 diabetes incidence in the canton of Zurich in Switzerland. The nature of the data is point process data, so I choose to fit a log-Gaussian Cox process with R-INLA. I select the stochastic partial differential equation (SPDE) approach to approximate the continuous spatial field.
Note: All the above analysis are subject of corrections and improvements. Feel free to contact me if you have further suggestions.
Faculty courses
- Advanced regression. 2023-now. (MSc Epidemiology) Imperial College London.
- Bayesian and spatial analysis. 2021-2022. (MSc Epidemiology) Imperial College London (with Prof Marta Blangiardo, Dr Monika Pirani and Dr Tullia Padellini).
- Spatial Analysis course. 2020. (MSc Epidemiology) Imperial College London (with Dr Monika Pirani and Dr Areti Boulieri).
Short courses
I have taught several different statistical courses in the past, some examples include:
- Bayesian Modeling for Environmental Health Workshop. 2023. Columbia University, New York.
- Spatial and spatio-temporal models with NIMBLE (2-days online course for TIES).
- Summer School at the University of Padova: Spatiotemporal models in Environmental Epidemiology. 2021.
- Bayesian statistics with R-INLA. 2018. University of Zurich.