openVA (version 1.0.9)

ConvertData.phmrc: Convert standard PHMRC data into binary indicator format

Description

The PHMRC data and the description of the format could be found at http://ghdx.healthdata.org/record/population-health-metrics-research-consortium-gold-standard-verbal-autopsy-data-2005-2011. This function convert the symptoms into binary indicators of three levels: Yes, No, and Missing. The health care experience (HCE) and free-text columns, i.e., columns named "word_****", are not considered in the current version of data conversion.

Usage

ConvertData.phmrc(input, input.test = NULL, cause = NULL,
  phmrc.type = c("adult", "child", "neonate")[1], cutoff = c("default",
  "adapt")[1], ...)

Arguments

input

standard PHMRC data format

input.test

standard PHMRC data format to be transformed in the same way as input

cause

the column name for the cause-of-death variable to use. For example, "va34", "va46", or "va55". It is used if adaptive cut-offs are to be calculated for continuous variables. See below for details.

phmrc.type

which data input format it is. The three data formats currently available are "adult", "child", and "neonate".

cutoff

This determines how the cut-off values are to be set for continuous variables. "default" sets the cut-off values proposed in the original paper published with the dataset. "adapt" sets the cut-off values using the rules described in the original paper, which calculates the cut-off as being two median absolute deviations above the median of the mean durations across causes. However, we are not able to replicate the default cut-offs following this rule. So we suggest users to use this feature with caution.

...

not used

Value

converted dataset with only ID and binary symptoms. Notice that when applying this function to the raw PHMRC data, the returned ID variable corresponds to the row index of the raw PHMRC data (i.e., cleaned data with ID = 10 correspond to the 10th row of the raw dataset), and does not correspond to the "newid" column in the PHMRC data.

References

James, S. L., Flaxman, A. D., Murray, C. J., & Population Health Metrics Research Consortium. (2011). Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies. Population Health Metrics, 9(1), 1-16.

Examples

# NOT RUN {
# read the raw data files from PHMRC website
# notice reading directly from internet could be time consuming
# so we only read 100 rows here.
# in practice, it is much easier and faster to download the file first,
#	and read all at once.
raw <- read.csv(getPHMRC_url("adult"), nrows = 100)
head(raw[, 1:20])
# default way of conversion
clean <- ConvertData.phmrc(raw, phmrc.type = "adult")
head(clean$output[, 1:20])
# using cut-offs calculated from the data (caution)
clean2 <- ConvertData.phmrc(raw, phmrc.type = "adult", 
						cause = "va55", cutoff = "adapt")
head(clean2$output[, 1:20])

# Now using the first 100 rows of data as training dataset
# And the next 100 as testing dataset
test <- read.csv(getPHMRC_url("adult"), nrows = 200)
test <- test[-(1:100), ]

# For the default transformation it does matter
clean <- ConvertData.phmrc(raw, test, phmrc.type = "adult")
head(clean$output[, 1:20])
head(clean$output.test[, 1:20])
# For adaptive transformation, need to make sure both files use the same cutoff
clean2 <-ConvertData.phmrc(raw, test, phmrc.type = "adult", 
						cause = "va55", cutoff = "adapt")
head(clean2$output[, 1:20])
head(clean2$output.test[, 1:20])
# }