A Bayesian Approach to Modelling Subnational Spatial Dynamics of Worldwide Non-State Terrorism, 2010 - 2016
A. Python, J. Illian, C. Jones-Todd, M. Blangiardo
J. R. Statist. Soc. A, XX (2018), XX -- XX

All results are obtained using R statistical software. We provide the R code and data needed to reproduce the research work.
For full replication, the R code listed below should be run in this order: 
-first: from (1) to (2): generate input data for the statistical models;
-second: (4), (5) and (6): the L,S, and F model, respectively;
-third: (7): the results and plots of the L,S, and F model, respectively.

Since fitting the models are computationally intense, we also provide the outputs of the models as .RData files along with the R code used to produce them.
The R code and data are provided in "R_code.zip", which contains the following folders:

(1) GTD
GTD.R: extraction of terrorism data from the Global Terrorism database (GTD)
Input: GTDsource.csv (file downloaded Jan 2017 from GTD) 
Output: GTDworld.csv used in R code (2): 

(2) DATA
Data_paper.R: extraction of covariate data based on GTD events locations
Input: covariates data with different formats in covariate.zip 
Output: paper_data.RData used in R code (4) to (9)

(3) FUNCTIONS
functions.R: functions to facilitate running the spatio-temporal models in R-INLA
used in in R code (4) to (9)

(4) L-MODEL
L_model.R: the Bernoulli space-time models of the lethality of terrorism
Inputs: paper_data.RData (2) and functions.R (3)
Outputs: 
-models with different sets of covariates: L0.Rdata to L7.RData ;
-selected final model: L7.Rdata
-models for plotting: L7.Rdata
-robustness test models: Lrob1.Rdata, Lmesh2.RData, Lmesh3.RData
-predictive models with different sets of covariates: L7pred.Rdata, L0pred.Rdata

(5) S-MODEL
S_model.R: the Poisson space-time models of the severity of lethal terrorism
Inputs: paper_data.RData (2) and functions.R (3)
Outputs: 
-models with different sets of covariates: S0.Rdata to S7.RData ;
-selected final model: S4.Rdata
-models for plotting: S4plot.Rdata
-robustness test models: bernrob1.Rdata, bernmesh2.RData, bernmesh3.RData
-predictive models with different sets of covariates: bern0pred.Rdata, bern4pred.Rdata

(6) F-MODEL 
F_model.R: the Poisson space-time models of the frequency of lethal terrorism
Inputs: paper_data.RData (2) and functions.R (3)
Outputs:
-models with different sets of covariates: F0.Rdata to F7.RData ;
-selected final model: Ffinal.Rdata
-models for plotting: F3plot.Rdata
-robustness test models: Frob1.Rdata, Ffinalagg075.RData, Ffinalagg025.RData,Ffinalagg100.RData, Ffinalagg150.RData
-predictive models with different sets of covariates: F0pred.Rdata, F6pred.Rdata

(7) RESULTS
-Results_Lmodel.R: generate the results of the L-models and the graphics in the manuscript
Inputs: paper_data.RData (2), functions.R (3), L-model outputs (4)
Outputs: plots and figures

-Results_Smodel.R: generate the results of the S-models and the graphics in the manuscript
Inputs: paper_data.RData (2), functions.R (3), S-model outputs (5)
Outputs: plots and figures

-Results_Fmodel.R: generate the results of the F-models and the graphics in the manuscript
Inputs: paper_data.RData (2), functions.R (3), F-model outputs (6), and country.shp
Outputs: plots and figures


Contact (first author):
Dr Andre Python
Malaria Atlas Project | University of Oxford
Big Data Institute | Li Ka Shing Centre for Health Information and Discovery
Old Road Campus | Headington | Oxford | OX3 7LF | United Kingdom
E: andre.python@bdi.ox.ac.uk
W: www.map.ox.ac.uk | www.bdi.ox.ac.uk