The DHS Program User Forum: India » Discrepancy in resident status between individual files and merged household file

Home » Countries » India » Discrepancy in resident status between individual files and merged household file

Show: Today's Messages :: Show Polls :: Message Navigator

Discrepancy in resident status between individual files and merged household file [message #24275]

Mon, 11 April 2022 06:48

desktop
Messages: 2
Registered: April 2022

Member

Hi,

After merging the individual questionnaires with the household member (PR) datasets per Tom Pollum's response in this thread ( https://userforum.dhsprogram.com/index.php?t=msg&th=6693 &start=0&)

, I noticed that the usual versus visiting residents differed between hv102 and (m)v135. See the R code below.

Discrepancies between women (1 = Usual, 2 = Visitor) and merged (PR+IR+MR) dataset (0 = Visitor, 1 = Usual)

table(women$V135, combined$HV102[combined$HV104 == 2], useNA = "ifany")
   
         0      1
  1  21537 655926
  2    686  21537

Discrepancies between men (1 = Usual, 2 = Visitor) and merged (PR+IR+MR) dataset (0 = Visitor, 1 = Usual)

table(men$MV135, combined$HV102[combined$HV104 == 1], useNA = "ifany")
   
         0      1
  1   1884 108320
  2     34   1884

Have I missed something, or are these discrepancies due to (m)v135 being reported by the individual themselves and hv102 being reported for all members by one person?

[Updated on: Mon, 11 April 2022 06:58]

Report message to a moderator

Re: Discrepancy in resident status between individual files and merged household file [message #24280 is a reply to message #24275]

Tue, 12 April 2022 10:58

desktop
Messages: 2
Registered: April 2022

Member

After cross-referencing my merge in R with what Tom did in STATA, I noticed several errors. Residency now checks out. Concatenating variables from the men's and women's questionnaire (such as (M)V35) has to be done after the datasets have been merged.

Below is the R code for anyone that wants to merge IR+MR+PR and does not have access to STATA.

# Import women's questionnaire
women <- read_sav("Your data location",
                  col_select = c("V001", "V002", "V003", "V005", "V135")

# Change colnames to match household members (PR) dataset
colnames(women)[which(names(women) == "V001")] <- "HV001"
colnames(women)[which(names(women) == "V002")] <- "HV002"
colnames(women)[which(names(women) == "V003")] <- "HVIDX"

#Sort by
attach(women)
women <- women[order(HV001, HV002, HVIDX), ]
detach(women)

men <- read_sav("Your file location",
                col_select = c("MV001", "MV002", "MV003", "MV005", "MV135"))

#Change colnames to match household members (PR) dataset
colnames(men)[which(names(men) == "MV001")] <- "HV001"
colnames(men)[which(names(men) == "MV002")] <- "HV002"
colnames(men)[which(names(men) == "MV003")] <- "HVIDX"

#Sort by
attach(men)
men <- men[order(HV001, HV002, HVIDX), ]
detach(men)

household <- read_sav("Your file location",
                      col_select = c("HV001", "HV002", "HVIDX", "HV005", "HV104", "HV027", "HV102"))

attach(household)
household <- household[order(HV001, HV002, HVIDX), ]
detach(household)

irpr <- merge(household, women, by = c("HV001", "HV002", "HVIDX"), all.x = T)

attach(irpr)
irpr <- irpr[order(HV001, HV002, HVIDX), ]
detach(irpr)

combined <- merge(irpr, men, by = c("HV001", "HV002", "HVIDX"), all.x = T)

# Weights
combined <- combined %>%
  mutate(weight = case_when(HV104 == 1 ~ MV005,
                            HV104 == 2 ~ V005))

# Re-weight men due to 15% sampling probability
combined <- transform(combined, adj_weight=ifelse(HV104 == 1 & HV027 == 1, weight*(1/.15),
                                                  weight))

combined <- combined %>%
  mutate(resident = case_when(HV104 == 1 ~ MV135,
                            HV104 == 2 ~ V135))

combined <- combined %>%
  mutate(resident = case_when(resident == 1 ~ 1,
                              resident == 2 ~ 0))

table(combined$resident, combined$HV102)
      0      1
  0  24141      0
  1      0 787667

all.equal(as.numeric(combined$HV102)[!is.na(combined$V005) | !is.na(combined$MV005)], combined$resident[!is.na(combined$resident)]
)

TRUE

Still some minor discrepancies for other variables though, such as marital status. More NAs in the PR file. Better to use variables in individual files, when possible?

#Add S301/SM213/HV116 to col_select calls for IR/MR/PR datasets to code in previous chunk

combined <- combined %>%
  mutate(marriage = case_when(HV104 == 1 ~ SM213,
                              HV104 == 2 ~ S301))

combined$marriage
Labels:
 value                        label
     0                Never married
     1            Currently married
     2 Married, gauna not performed
     3                      Widowed
     4                     Divorced
     5                    Separated
     6                     Deserted

combined$HV116
Labels:
 value                 label
     0         Never married
     1     Currently married
     2 Formerly/ever married

table(combined$marriage, combined$HV116)
   
         0      1      2
  0 207332   2198    265
  1   1892 566533   1402
  2   1718    499     36
  3    106   1114  20034
  4    113    220   3126
  5     70    634   3406
  6     16    109    938

sum(table(combined$marriage))-sum(table(combined$HV116[!is.na(combined$V005) | !is.na(combined$MV005)]))
[1] 47

[Updated on: Tue, 12 April 2022 11:01]

Report message to a moderator

Previous Topic:	Merging Data DHS 2015-2016
Next Topic:	NFHS-4 issue with district variable

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Thu Nov 28 09:01:25 Coordinated Universal Time 2024