Home » Countries » Nigeria » Inquiry regarding data merging with DHS 2018 using R
| Inquiry regarding data merging with DHS 2018 using R [message #26109] |
Wed, 08 February 2023 22:23  |
woojae1995
Messages: 6
Registered: January 2023
|
Member |
|
|
I am currently analyzing Nigeria 2018 DHS dataset for a malaria project. I am using the household data (HR) & household member data (PR)
Because I am interested in ITN data, I referred to the github file: https://github.com/DHSProgram/DHS-Indicators-R/blob/ed302f3e 5afc73e44d6113a64183173d725f8fbd/Chap12_ML/ML_EXISTING_ITN.R to produce indicators for ITN
However, I am struggling with data merging / changing household data to long format and then filtering it to only include data for children of age 6-59 months which is the scope of my research
After conducting this part of reshaping the dataset presented from the above URL;
# Reshaping the dataset to a long format to tabulate among nets
myvars <- c(paste("hhid"),
paste("hml10_", 1:7, sep = ""))
HRdata_long1 <- reshape::melt(as.data.frame(HRdata[myvars]), id = c("hhid"))
HRdata_long1$idx <- str_sub(HRdata_long1$variable,-1,-1)
HRdata_long1$variable <- NULL
names(HRdata_long1)[names(HRdata_long1) == c("value")] <- c("hml10")
myvars <- c(paste("hhid"),
paste("hml21_", 1:7, sep = ""))
HRdata_long2 <- reshape::melt(as.data.frame(HRdata[myvars]), id = c("hhid"))
HRdata_long2$idx <- str_sub(HRdata_long2$variable,-1,-1)
HRdata_long2$variable <- NULL
names(HRdata_long2)[names(HRdata_long2) == c("value")] <- c("hml21")
HRdata_long <- merge(HRdata_long1,
HRdata_long2, by = c("hhid", "idx"))
myvars <- c("hhid","hv005","hv025", "hv024", "hv270")
HRdata_long3 <- (as.data.frame(HRdata[myvars]))
HRdata_long <- merge(HRdata_long,
HRdata_long3, by = c("hhid"))
I get an exploded number of data entries of more than 200000 compared to something around 40,000 in the original HR data. I presume this is because by these commands the household data was expanded for all household members (1 household data -> n household members data)
But my question is; how do I select only children 6-59 months from this data?
I have tried the following for the last 3 lines of code;
myvars <- c("hhid","hv005","hv025", "hv024", "hv270", "hv014", "hc1")
HRdata_long3 <- (as.data.frame(HRdata[myvars]))
HRdata_long <- merge(HRdata_long,
HRdata_long3, by = c("hhid"))
HRdata_long <- filter(HRdata_long, hc1>=6, hc1 <=59)
But this will still give me 268121 observations for HRdata_long, which is far greater than what I got from my previous coding I identified. It was 11590 observations for HR data when restricted to children of 6~59 months by using the PR data as the following:
# keep relevant vars
PRtemp =subset(PRdata, select=c(hv001, hv002, hc1), 'NA'= TRUE)
#perform merge
HRdata <- merge(HRdata,PRtemp,by=c("hv001", "hv002"))
HRdata <- filter(HRdata, hc1 >=6, hc1 <=59)
rm(PRtemp)
Can anyone explain me the differences between the two approach and why I can't restrict the above long-format to just children?
From my guess, I think it is because when expanded to the long format, even for household members that are not children, they will still have the hc1 variable between 6-59 as long as they had a children of 6-59 months age in their original household. Is this the case?
If so, how can I work around to only restrict to the actual children data when I want to expand the HR data to the long format?
|
|
|
|
| Re: Inquiry regarding data merging with DHS 2018 using R [message #26137 is a reply to message #26109] |
Fri, 10 February 2023 11:59  |
Janet-DHS
Messages: 938
Registered: April 2022
|
Senior Member |
|
|
Following is a response from DHS staff member, Tom Pullum:
You posted two related emails but I hope this response will help with both of them.
The PR file has one record for each household member. It is just a long version of the HR file, which has one (very wide) record for each household. If you want to select certain types of household members, in your case children age 6-59 months, you do this with the PR file. Each record in the PR file retains the information about the household as a whole, such as wealth quintile (hv270).
The child's age in months is given by hc1 in the PR file. You would select the cases with hc1>=6 & hc1<=59. To repeat, no merging is needed to get these cases.
|
|
|
|
Goto Forum:
Current Time: Fri Oct 31 13:12:46 Coordinated Universal Time 2025
|
Members
Search
Help
Register
Login
Home
