| Reading data files into R Studio [message #11707] |
Thu, 02 February 2017 06:28  |
DHS user
Messages: 111
Registered: February 2013
|
Senior Member |
|
|
I am conducting research using the 1990 & 2010 Pakistan DHS. I am interested in utilizing the household and woman files to understand ways in which factors such as education, income, religion, age at marriage etc. impact a woman's family planning uptake / contraceptive use.
I have dowloaded the FLAT files for the 1990 and 2012 PDHS and there are a few issues that are preventing me from reading data files into RStudio. As an example, for file "PKBR21FL.DAT" I examined the first two rows of this data file.
1. The rows do not appear to have the same number of elements. I presume there are missing entries which might be causing this. I checked in the "Coding Standards" section of DHS VI Individual recode manual (pdf), which mentions that a value of BLANK means "Variable is not applicable for this respondent either because the question was not asked in a particular country or because the question was not asked of this respondent due to the flow or skip pattern of the questionnaire."
Without knowing how to parse the rows of the file, a statistical software program such as R/RStudio cannot read this into a data matrix of fixed size, with rows corresponding to the individual records in the flat file, and columns corresponding to all possible individual parameters/ factors. Since the data files are not comma delimited, I do not know how to proceed with parsing missing data for individuals that have been replaced with a BLANK. Does the latter mean " "?
2. It is very unclear as to what order the columns in "PKBR21FL.DAT" are in. I cannot find in the PDF manual a place where it provides a mapping between the ordering of the columns, and their definitions as they appear on page 9 (Section H00). As an example, the Country variable appears in position 6, however this appears as variable #2 (HV000) on page 9.
Your clarification of these two questions would be greatly appreciated.
|
|
|
|
| Re: Reading data files into R Studio [message #11708 is a reply to message #11707] |
Thu, 02 February 2017 06:31  |
Bridgette-DHS
Messages: 3208
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Specialist, Trevor Croft:
The data file you are looking at is a fixed format file, in which columns of the records determining which variable is which. Each record should have the same length, but in some files records with trailing blanks are truncated. You can find the layout of these records by looking at any of the .DCT (for Stata), .SAS (for SAS), or .SPS (for SPSS) files. These are all text files that describe the layout of the data, and you can use this information to construct code to read the data into R.
However, the easiest way to get data into R is actually to start with either the Stata or SPSS datasets. I generally prefer the Stata dataset, but they both work. You can use the read.dta() function, as follows using the Stata dataset:
dta <- read.dta("PKBR21FL.dta", convert.factors = FALSE)
read.dta() is in the package "foreign", so you will need
install.packages("foreign")
library(foreign)
I prefer not to convert variables to factors automatically so I use convert.factors = FALSE, but you may prefer to have it set to TRUE and automatically convert. If you don't automatically convert variables to factors, then you can use code such as
dta$sex <-factor(recode(dta$b4,"1='1 Male';2='2 Female';9='9 Missing';else=NA"))
or even
dta$sex = factor(dta$b4)
|
|
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
