anonymous-submission-acl2025's picture
add 27
3af64c0
#Table 4: Comparison of most common occupations for network and household data sets.
#install packages
# install.packages('plyr')
# install.packages('dplyr')
# install.packages('tidyr')
# install.packages('ggplot2')
# install.packages('multiwayvcov')
# install.packages('lmtest')
# install.packages('stargazer')
rm(list=ls())
#setwd() #set working directory
library(plyr);library(dplyr, warn.conflicts = FALSE)
library(tidyr);library(ggplot2)
suppressMessages(library(multiwayvcov, warn.conflicts = F))
suppressMessages(library(lmtest, warn.conflicts = F))
suppressMessages(library(stargazer))
s = function(x){summary(factor(x))}
#Read network data
A = readRDS('4-20-20_Network-KNearest_DeID_demed.RDS')
#Read and clean household sample data
H = read.csv('4-20-20_deid_nearestK.csv',
na.strings=c('','NA'),strip.white=T,stringsAsFactors = F)
Hprime = H[H$A.A7_Area.Neighborhood %in% names(which(table(H$A.A7_Area.Neighborhood) >= 30)),]#Drop neighborhoods with < 30 observations
Hprime = Hprime[which(!is.na(Hprime$A.A7_Area.Neighborhood)),] #drop NA neighborhoods
Hprime = Hprime[which(Hprime$A.A7_Area.Neighborhood != 'Bangalore NA'),]
H = Hprime; rm(Hprime)
########################################################################################################################
############################################################################################################
#Map values to occupations
A$Job = mapvalues(A$D.D1_Occupation,
from = c(12,
13,
19,
2,
20,
21,
3,
4,
6,
7,
9),
to = c('Garbage',
'Gardener',
'Security',
'Butcher',
'Tailor',
'Vendor',
'Carpenter',
'Construction',
'Cook',
'Corporate',
'Electrician'))
H$Job = mapvalues(H$D.D1_Occupation.,
from = c(1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26, #B17: retired
27), #B17: unemployed
to = c('Agriculture',
'Butcher',
'Carpenter',
'Construction',
'Labour',
'Cook',
'Corporation',
'Driver',
'Electrical',
'Factory',
'Flower',
'Garbage',
'Gardener',
'Maid',
'Mechanic',
'Painter',
'ProfessionalSvc',
'Grocessory',
'Security',
'Tailor',
'Vendor',
'Government',
'Housewife',
'Student',
'Other',
NA, NA )) #Map retired/unemployed to NA so they don't affect denominator
#Note that B17 includes retired/unemployed; these are mapped above to NA and dropped from table (and not included in denominator)
#Network
sort (s(A$Job) / sum(!is.na(A$Job )) * 100 , decreasing = T) %>% round(2)
#Household
sort (s(H$Job) / sum(!is.na(H$Job )) * 100 , decreasing = T) %>% round(2)
#These are basis for manually-created Table 4
################################################################################################