In alekrutkowski/wiod.diagrammer: Easy work with WIOD (the 2016 release) including diagramming (flowcharts)

# Rendering this doc:
# rmarkdown::render(paste0(getwd(),'/README.Rmd'), envir=globalenv())
knitr::opts_chunk$set(echo = TRUE)
knitr::opts_chunk$set(engine = 'R')

Installation

## if package `devtools` not installed, first do this:
# install.packages('devtools')
devtools::install_github('alekrutkowski/wiod.diagrammer')

Documentation

https://alekrutkowski.github.io/wiod.diagrammer/reference/index.html

Example

Load WIOD data for a given year from an official WIOD .Rdata file (release 2016), which has to be first downloaded from http://www.wiod.org/protected3/data16/wiot_ROW/wiot_r_Nov16.zip and extracted (unzipped).

library(wiod.diagrammer)

library(magrittr)
plotLinks <- function(...) # a trick to plot svg
    wiod.diagrammer::plotLinks(...) %>% 
    DiagrammeRsvg::export_svg() %>% 
    cat(file='Graph.svg')
print <- function(d) # Just for clarity
    `if`(d %>% is.data.frame,
         base::print.data.frame(`colnames<-`(d, paste0(colnames(d),'  ')),
                                right=FALSE,
                                row.names=FALSE),
         base::print.default(d))

W <- loadWIOD('WIOT2014_October16_ROW.RData')

# Check the names of the first 10 columns:
cat(head(colnames(W), 20),
    sep='\n')
# How many columns and rows?
message(ncol(W),' columns; ',nrow(W),' rows')

Now let's flatten (reshape into long format):

W_flat <- flatWIOD(W)
# See the structure of W:
str(W_flat)
head(W_flat, 10)
tail(W_flat, 10)

wiod.diagrammer uses internally the data.table package for performance. You can work with W_flat using the data.table's semantcs (e.g. the := in-place column generation/modification) or, if you prefer base R data.frames, you may convert W_flat into a regular data.frame with the function as.data.frame, do the modifications, and then convert it back into data.table with the function data.table::as.data.table for further processing (through the function findLinks described below).

Now let's find the top supplier-user linkages in WIOD (as defined by the column value in W_flat which -- if not modified -- corresponds to the intermediate consumption in W).

First let's introduce a helper function tieRobustRankLessOrEqual (available in wiod.diagrammer) comparing it with base R rank:

x <- c(1,1,2,2,2,3,3)
# A comparison:
print(data.frame(x,
                 rank(x) <= 2,
                 tieRobustRankLessOrEqual(x, 2),
                 check.names = FALSE))

Now, let's find e.g. the 3 main customers of German automotive industry as well as 2 main customers of those customers and 1 main customer of those customer's customers (we could go even deeper if we wanted or use different cut-off ranks, not necessarily in the declining order). So, we have 3 levels of linkages (the 1st one -- direct, the 2nd one and the 3rd one -- indirect). NB: Some industries may re-emerge at different rounds and they may be both customers and suppliers at the same time. Let's also differentiate our ranks (top 3, 2, and 1) by one more dimension: domestic vs foreign linkages.

W_flat[, domestic :=   # creating column in-place, following data.table's semantics
           ExpCountry == ImpCountry]
# Let's get rid of self-produced intermediate consumption
W_flat_noself <- W_flat[!(domestic &
                              ExpSectorNr == ImpSectorNr)]
# Let's keep only intra-EU trade
COUNTRIES_DT <- countries()
EU_COUNTRIES <-
    COUNTRIES_DT$Country[COUNTRIES_DT$isEUmember]
# Let's keep only flows >= 1 billion USD for clarity
W_flat_noself_truncated <- W_flat_noself[value >= 1000 &  # original WIOD data is in million USD
                                             ExpCountry %in% EU_COUNTRIES &
                                             ImpCountry %in% EU_COUNTRIES]
TOP_CUSTOMERS <-
    findLinks(partners = 'users',
              flat_wiod = W_flat_noself_truncated,
              start_countries = 'DEU', # We could add here other countries.
              start_sectors = 20, # "Manufacture of motor vehicles, trailers and semi-trailers"
              # We could add here other sectors.
              ListOfselectionFuns = # As discussed in the text above:
                  list(function(flat_wiod) tieRobustRankLessOrEqual(-flat_wiod$value, 3),  # minus because we
                       function(flat_wiod) tieRobustRankLessOrEqual(-flat_wiod$value, 2),  # want to rank from
                       function(flat_wiod) tieRobustRankLessOrEqual(-flat_wiod$value, 1)), # the highest to the lowest
              by = c('domestic','ExpCountry','ExpSectorNr')) # as discussed above
str(TOP_CUSTOMERS)
head(TOP_CUSTOMERS, 20)

Now let's do a similar exercise, just "upstream" i.e. for suppliers of suppliers.

TOP_SUPPLIERS <-
    findLinks(partners = 'suppliers',
              flat_wiod = W_flat_noself_truncated,
              start_countries = 'DEU', # We could add here other countries.
              start_sectors = 20, # "Manufacture of motor vehicles, trailers and semi-trailers"
              # We could add here other sectors.
              ListOfselectionFuns = # As discussed in the text above:
                  list(function(flat_wiod) tieRobustRankLessOrEqual(-flat_wiod$value, 3),  # minus because we
                       function(flat_wiod) tieRobustRankLessOrEqual(-flat_wiod$value, 2),  # want to rank from
                       function(flat_wiod) tieRobustRankLessOrEqual(-flat_wiod$value, 1)), # the highest to the lowest
              by = c('domestic','ImpCountry','ImpSectorNr')) # as discussed above
str(TOP_SUPPLIERS)
head(TOP_SUPPLIERS, 20)

Now, let's plot the "upstream" linkages, making all the German sectors blue. By default, the cross-border flows are dashed, while the domestic flows are solid lines (arrows). The numbers in the nodes (rectangles) represent, by default, the sector output (or the total intermediate consumption for the final use sectors such as final consumption or investment if they show up in the customers' graphs).

In wiod.diagrammer the rendering of the plot is done internally by DiagrammeR::grViz. The plots can be saved manually in RStudio, or programmatically e.g. to an .svg file via DiagrammeRsvg::export_svg to a character vector and then cated, or to a .png file piping them through DiagrammeRsvg::export_svg, charToRaw and rsvg::rsvg_png.

plotLinks(top_links_dt = TOP_SUPPLIERS,
          wiot = W, # this is necessary
          specificNodeOptionsFun =  # this is optional, just to show-off:
            function(country_sector_dt)
                ifelse(country_sector_dt$Country=='DEU',
                       'style=filled, fillcolor=cadetblue1', "")) # GraphViz colour names can be found at:
                                                                  # http://www.graphviz.org/doc/info/colors.html

Click on the picture to zoom in:

What else is available in the package? The functions which produce auxiliary data.tables (that are used by wiod.diagrammer's plotLinks function if evaluated with default argument values).

COUNTRIES <- countries() # NB: no argument to function `countries`
str(COUNTRIES)
print(COUNTRIES)
SECTORS <- sectors(W)
SECTORS[, SectorLab :=  # truncate the long sector labels just for clarity below
            substr(SectorLab, 1, 30)]
str(SECTORS)
head(SECTORS, 20)
AGGREGATES <- aggregates(W)

The variable/column names produced by the function aggregates reflect those in WIOD. They are explained in the documentations of aggregates.

str(AGGREGATES) 
head(AGGREGATES, 20)

alekrutkowski/wiod.diagrammer documentation built on May 20, 2019, 8:47 a.m.