Data Description and Explanation
for the SMJ publication:

##################################################################################
#### Exploring Dependence Structures in the International Arms Trade Network: ####
##################### A Network Autocorrelation Approach #########################
##################################################################################

by

Michael Lebacher
Department of Statistics, Faculty of Mathematics, Informatics and Statistics,
Ludwig-Maximilians-Universitat Munchen, Germany
michael.lebacher@stat.uni-muenchen.de

and

Paul W. Thurner
Geschwister Scholl Institute of Political Science, Faculty of Social Scicences,
Ludwig-Maximilians-Universitat Munchen, Germany

and

Gran Kauermann
Department of Statistics, Faculty of Mathematics, Informatics and Statistics,
Ludwig-Maximilians-Universitat Munchen, Germany


##################################################################################
################# Data Descprition ###############################################
##################################################################################

The Data is supplied in the file Data.RData and contains the following elements:

### amk
A list of 67 weighted adjacency matrices of dimension (257x257). The first matrix
represents the year 1950 (amk[[1]]) and the last one the year 2016 (amk[[67]]). 
Within each rows are the exports of the country denoted in the row to the importers
in the columns. I.e. amk[[1]][201,202] is the transfer of United Kingdom 
(country index=201) to the United States (country index=202) in 1950 measured in 
TIV values

### laenderliste
A dataframe of dimension (257X6) giving the names and abbrevations of the countries
in the same ordering as the adjacency matrices. Mainly needed for conversion of
different country codes (name vs. ISO3)

### nEX
A binary matrix of dimension (257x67) with the countries in the rows and a binary 
indicator of inclusion in the model in the columns. I.e. a zero in nEX[154,1] says
that Russia (country code 154) is not included in the model in 1950 while nEX[176,1]=1
means that the Soviet Union (country code 176) is to be included in 1950.

### R
A (18964x2) dimensional matrix encoding for all 18964 modelled transfers the country code
of the exporting and importing country. I.e. R[1,]=c(11,135) gives a transfer from 
Australia (country code=11) to New Zealand (country code=135).

### TC1, TC2, TC3
Lists of dependence structures for the Spatial weights of varying dimension. The
function that creates them from the networks is provided in the R-file Simulations.R in 
the folder 4) Simulations.

### X
A (18964x8) matrix containing the time (t), the exports in TIV (EX_ij), the logarithmic
GDP of the exporting and importing country (lGDP_i, lGDP_j), the logarithmic spatial distance
between them (D_ij), a binary indicator for formal alliances (DA_ij) and the absolute
difference of the polity IV score (POL_ij).

### Z
A (18964x448) matrix that encodes the information of R as a matrix for the 224 most
important countries with the first 1:224 entries being the exporter and the following
225:448 entries being the importer. I.e. Z[1,] contains only zeros except for the entry
11 Australia (country code=11) and 359 New Zealand (country code=359-224=135).


### names
A (1x257) vector of country names.

### relevant
The same list as amk with the difference that here transfers not considered in
nEX are excluded

### Multiplot
A function that allows plotting multiple ggplots in one plot.


##################################################################################
################# Workflow Descprition ###########################################
##################################################################################

### 1) Drawing descriptive Figures
Contains a R file (descriptive_figures.R) that allows to reproduce the descriptive Figures of the paper.

### 2) Estimating the Models
Contains a R file (Estimation_and_AIC.R) that allows for estimating all four regressions (Linear Model
and different weight matrices) and the calculation of the respective AIC. It also reproduces the respective
AIC tables of the main paper and the supplementary material, the plot with the coefficients.
Additionally in the file (Residuals.R) the Figures related to the residual analysis are included.

### 3) Supplementary Material contains a R file (Supplementary_Material.R) that allows to reproduce the
results of the analysis with multiple weight matrices simultaneously, additionally it containes the R file
Supplementary_Material2.R which allows to compare the SEM with the SAR model as well as the model-averaged
version of the model.

### 4) Simulations
Contains the function that creates TC1, TC2 and TC3 as well as the procedure to build them up for all years.
After the simulation of the networks, they are re-estimated as described in the paper.
