2.1 DHS dataset types

The table below lists the different types of DHS data and their naming conventions in surveyPrev. You can find more details in this website

2.2 Option 1: Specifying data processing rules

For many standard indicators, the specification can be summarized into the following three steps:

1. Filter the data for relevant individuals based on certain selection criterion.
2. Specify a yes response, i.e., outcome value = 1, if certain conditions hold.
3. Specify a no response, i.e., outcome value = 0, if certain conditions hold.

If an indicator can be specified in this way, with the conditions depending on variables in the DHS data, they can be specified using the filter, yesCondition, and noCondition arguments in the getDHSindicator() function.

As an example, we create a dataset below that consists of neonatal deaths during the last 5 years prior to survey. The dataset we need is the births recode. After obtaining the raw data using getDHSdata() function, we specify the following conditions:

1. Filter the dataset with v008 - b3 < 120. v008 and b3 are column names in the DHS birth recode. v008 is the data of interview, and b3 is the date of birth. Both dates are in Century Month Code (CMC) format, so that the difference is in the unit of months. This filter creates a dataset that only consist of births 120 months prior to survey.
2. Specify a yes response, i.e., a neonatal death, by b7 == 0. b7 is the column specifying age of deaths. Age of deaths equal to 0 means the death happened during the first month of birth.
3. Specify a no response, i.e., not a neonatal death, by b7 > 0 | is.na(b7). That is, deaths after the first month or live children are considered not a neonatal death.

Putting everything together, the dataset can be created with the following codes.

Recode <- "Births Recode"
dhsData <- getDHSdata(country = "Zambia", indicator = NULL, 
                        Recode = Recode, year = "2018")
data <- getDHSindicator(Rdata = dhsData,
                        indicator = NULL, 
                        filter = "v008 - b3 < 120",
                        yesCondition = "b7 == 0", 
                        noCondition = "b7 > 0 | is.na(b7)")

Notice that all the conditions are specified by a string of expressions that R can use to filter and data with. For the filter argument, more than one filters can be applied. For example, the following two calls to getDHSindicator() are equivalent and add an additional filter for births by mothers over 30 years old at time of birth.

data_over30_a <- getDHSindicator(Rdata = dhsData,
                        indicator = NULL, 
                        filter = c("v008 - b3 < 120", "b3 - v011 > 30 * 12"),
                        yesCondition = "b7 == 0", 
                        noCondition = "b7 > 0 | is.na(b7)")
data_over30_b <- getDHSindicator(Rdata = dhsData,
                        indicator = NULL, 
                        filter = "v008 - b3 < 120 & b3 - v011 > 30 * 12",
                        yesCondition = "b7 == 0", 
                        noCondition = "b7 > 0 | is.na(b7)")

2.3 Option 2: Specifying function to process the indicator

When the indicator specification needs more steps than the first option can handle, users can import the full data processing functions by following the steps below.

Let’s take Current use of any modern method of contraception (all women) as an example. For users familiar with the standard indicators defined by the DHS Data Indicator API, the indicator ID is “FP_CUSA_W_MOD”. We will go through the steps to create the customized function below.

Step 1: Search indicator ID or key words in the Guide to DHS Statistics, and then identify which chapter it is in.

For our example, we can either search “FP_CUSA_W_MOD”, or “contraceptive”, and it is in chapter 7: family planning.

Figure 1: Screenshot of Step 1(a): Searching for indicator ID or key words.

Figure 2: Screenshot of Step 1(b): Identifying which chapter the indicator is from.

Step 2: We can download IndicatorList.xlsx from DHS GitHub site, and search keyword again in the the corresponding chapter to find out

1. which DHS data recode is used to create this indicator
2. which file contains the code to create this indicator in the DHS GitHub repository.
3. what the corresponding variable name is in the DHS GitHub repository.

For our example, since we are looking up for the indicator of all women currently using any modern method of contraception, we identify the cell “currently use any modern method”, and find out that the code to process this indicator is in the FP_USE.do file and we need IRdata (Individual Recode). We also identify the variable name used in the R codes on the Github repository is “fp_cruse_mod”. We will all three pieces of information in the next step to find the codes processing the indicator.

Figure 3: Screenshot of Step 2: Finding file name and recode name.

Step 3: In the Github repository, we find the folder for Chapter 7, the file FP_USE.R, and search “fp_cruse_mod” to find the following chunk of code script.

Figure 4: Screenshot of Step 3: Finding code.

We extract the following chunk of codes for this indicator, which takes the IR data, and perform a few steps of data cleaning.

# Currently use modern method
IRdata <- IRdata %>%
    mutate(fp_cruse_mod = ifelse(v313 == 3, 1, 0)) %>%
    set_value_labels(fp_cruse_mod = c(yes = 1, no = 0)) %>%
    set_variable_labels(fp_cruse_mod = "Currently used any modern method")

We can use the code chunk to define a new function to be used in the getDHSindicator function. The self-defined function should:

• Use Recode file as input and return the same data.frame.
• Change the name of your variable into "value" in the end.

The example below creates a fp_cruse_mod function for “Current use of any modern method of contraception (all women)”, which can be recognized by the getDHSindicator function later.

fp_cruse_mod <- function(RData) {
    IRdata <- RData %>%
        mutate(fp_cruse_mod = ifelse(v313 == 3, 1, 0)) %>%
        set_value_labels(fp_cruse_mod = c(yes = 1, no = 0)) %>%
        set_variable_labels(fp_cruse_mod = "Currently used any modern method")
    colnames(IRdata)[colnames(IRdata) == "fp_cruse_mod"] <- "value"
    return(IRdata)
}

Finally, after we create this function fp_cruse_mod, We can create the indicator by

1. Use getDHSdata function to downloading relevant DHS datasets using the identified DHS data type in Step 3. In this example we specify Recode = "Individual Recode" and indicator = NULL, The recode be one of the recode lists in table 1. to download the dataset. We can also set Recode = NULL, in which case all available DHS data types will be downloaded.
2. Use the function FUN = fp_cruse_mod in the call of getDHSindicator to process the indicator according to the customized function.

Altogether, the following codes creates the dataset of the processed indicator.

year <- 2018
country <- "Zambia"
Recode <- "Individual Recode"
dhsData <- getDHSdata(country = country, indicator = NULL, Recode = Recode, year = year)
data <- getDHSindicator(dhsData, indicator = NULL, FUN = fp_cruse_mod)
head(data)

##   cluster householdID    v024  weight strata value
## 1       1           1 eastern 1892890  rural     0
## 2       1           2 eastern 1892890  rural     0
## 3       1           3 eastern 1892890  rural     0
## 4       1           4 eastern 1892890  rural     1
## 5       1           7 eastern 1892890  rural     1
## 6       1           9 eastern 1892890  rural     1