In SWS-Methodology/hsfclmap: Mapping table between country-specific HS codes and FCL codes

knitr::opts_chunk$set(echo = TRUE)
library(stringr)
library(ggplot2)
library(dplyr, warn.conflicts = FALSE)

How many HS ranges in the map are wider than single HS6 code.

data("hsfclmap2", package = "hsfclmap")

hsfcl <- hsfclmap2 %>% 
  mutate(hs6from = str_extract(fromcode, "^\\d{1,6}"),
         hs6to   = str_extract(tocode, "^\\d{1,6}"))

hsfcl %>% 
  summarize(sum(hs6from != hs6to)/n())

How many HS-to-FCL mapping records have one-to-many link. We take into account only left-hand side of HS range.

fclinhs <- hsfcl %>% 
  select(mdbyear, area, validyear, flow, hs6from, fcl) %>% 
  group_by(mdbyear, area, validyear, flow, hs6from) %>% 
  mutate(fclinhs = length(unique(fcl))) %>% 
  ungroup()

# Total share
fclinhs %>% 
  summarise(sum(fclinhs > 1) / n())

# By year
fclinhs %>% 
  group_by(mdbyear) %>% 
  summarise(one2manyprop = sum(fclinhs >1) / n()) %>% 
  ggplot(aes(mdbyear, one2manyprop)) +
  geom_path(stat = "identity")

How many duplicate links are in the map if we don't take into account year of map production (mdbyear).

hsfclmap2 %>% 
  select(validyear, area, flow, fromcode, tocode, fcl) %>% 
  distinct() %>% 
  summarize(dupl_prop = 1 - n() / nrow(hsfclmap2))

So we can drop duplicates!

hsfcluniq <- hsfcl %>% 
  select(-starts_with("mdb")) %>% 
  distinct()

dim(hsfcluniq)

How many one-to-many HS6-to-FCL links now?

hsfcluniq %>% 
  select(validyear, area, flow, hs6from, fcl) %>% 
  group_by(validyear, area, flow, hs6from) %>% 
  summarize(fclnumb = length(unique(fcl))) %>% 
  ungroup() %>% 
  summarize(sum(fclnumb > 1) / n())

SWS-Methodology/hsfclmap documentation built on May 9, 2019, 11:53 a.m.