Health and Retirement Study (HRS)
The Health and Retirement Study interviews Americans aged 50+ for their entire life. Allows for findings like, “Among Americans who were 50-74 years old in 1998, X% lived in nursing homes by 2010.”
Many tables, most with one row per sampled respondent and linkable over time. Use the RAND HRS data file for a cleaner, cross-wave data set.
A complex sample survey designed to generalize to Americans aged 50+ at each interview, but longitudinal analysts can observe outcomes.
Released biennially since 1992.
Administered by the University of Michigan’s Institute for Social Research with data management by the RAND Corporation. Funded by the National Institute on Aging and the Social Security Administration.
Simplified Download and Importation
The R lodown
package easily downloads and imports all available HRS microdata by simply specifying "hrs"
with an output_dir =
parameter in the lodown()
function. Depending on your internet connection and computer processing speed, you might prefer to run this step overnight.
library(lodown)
lodown( "hrs" , output_dir = file.path( path.expand( "~" ) , "HRS" ) ,
your_username = "username" ,
your_password = "password" )
lodown
also provides a catalog of available microdata extracts with the get_catalog()
function. After requesting the HRS catalog, you could pass a subsetted catalog through the lodown()
function in order to download and import specific extracts (rather than all available extracts).
library(lodown)
# examine all available HRS microdata files
hrs_cat <-
get_catalog( "hrs" ,
output_dir = file.path( path.expand( "~" ) , "HRS" ) ,
your_username = "username" ,
your_password = "password" )
# RAND consolidated file only
hrs_cat <- subset( hrs_cat , grepl( 'rand([a-z]+)stata\\.zip' , file_name ) )
# download the microdata to your local computer
hrs_cat <- lodown( "hrs" , hrs_cat ,
your_username = "username" ,
your_password = "password" )
Analysis Examples with the survey
library
Construct a complex sample survey design:
library(survey)
hrs_df <-
readRDS(
grep(
'rand([a-z]+)stata(.*)rds$' ,
list.files(
hrs_cat$output_folder ,
full.names = TRUE ,
recursive = TRUE
) ,
value = TRUE
)
)
# RAM cleanup
keep_vars <-
c( "raehsamp" , "raestrat" , "r3wtresp" ,
"r3work" , "r12work" , "h12ahous" ,
"r3mstat" , "r12mstat" , "h4ahous" )
hrs_df <- hrs_df[ keep_vars ]
# community residents aged 50+ in 1996
hrs_design <-
svydesign(
id = ~ raehsamp ,
strata = ~ raestrat ,
weights = ~ r3wtresp ,
nest = TRUE ,
data = subset( hrs_df , r3wtresp > 0 )
)
Variable Recoding
Add new columns to the data set:
hrs_design <-
update(
hrs_design ,
one = 1 ,
working_in_1996 = r3work ,
working_in_2014 = r12work ,
marital_stat_1996 =
factor( r3mstat , levels = 1:8 , labels =
c( "Married" , "Married, spouse absent" ,
"Partnered" , "Separated" , "Divorced" ,
"Separated/divorced" , "Widowed" ,
"Never married" ) ) ,
marital_stat_2014 =
factor( r12mstat , levels = 1:8 , labels =
c( "Married" , "Married, spouse absent" ,
"Partnered" , "Separated" , "Divorced" ,
"Separated/divorced" , "Widowed" ,
"Never married" ) )
)
Unweighted Counts
Count the unweighted number of records in the survey sample, overall and by groups:
sum( weights( hrs_design , "sampling" ) != 0 )
svyby( ~ one , ~ marital_stat_1996 , hrs_design , unwtd.count )
Weighted Counts
Count the weighted size of the generalizable population, overall and by groups:
svytotal( ~ one , hrs_design )
svyby( ~ one , ~ marital_stat_1996 , hrs_design , svytotal )
Descriptive Statistics
Calculate the mean (average) of a linear variable, overall and by groups:
svymean( ~ h12ahous , hrs_design , na.rm = TRUE )
svyby( ~ h12ahous , ~ marital_stat_1996 , hrs_design , svymean , na.rm = TRUE )
Calculate the distribution of a categorical variable, overall and by groups:
svymean( ~ marital_stat_2014 , hrs_design , na.rm = TRUE )
svyby( ~ marital_stat_2014 , ~ marital_stat_1996 , hrs_design , svymean , na.rm = TRUE )
Calculate the sum of a linear variable, overall and by groups:
svytotal( ~ h12ahous , hrs_design , na.rm = TRUE )
svyby( ~ h12ahous , ~ marital_stat_1996 , hrs_design , svytotal , na.rm = TRUE )
Calculate the weighted sum of a categorical variable, overall and by groups:
svytotal( ~ marital_stat_2014 , hrs_design , na.rm = TRUE )
svyby( ~ marital_stat_2014 , ~ marital_stat_1996 , hrs_design , svytotal , na.rm = TRUE )
Calculate the median (50th percentile) of a linear variable, overall and by groups:
svyquantile( ~ h12ahous , hrs_design , 0.5 , na.rm = TRUE )
svyby(
~ h12ahous ,
~ marital_stat_1996 ,
hrs_design ,
svyquantile ,
0.5 ,
ci = TRUE ,
keep.var = TRUE ,
na.rm = TRUE
)
Estimate a ratio:
svyratio(
numerator = ~ h4ahous ,
denominator = ~ h12ahous ,
hrs_design ,
na.rm = TRUE
)
Subsetting
Restrict the survey design to :
sub_hrs_design <- subset( hrs_design , working_in_1996 == 1 )
Calculate the mean (average) of this subset:
svymean( ~ h12ahous , sub_hrs_design , na.rm = TRUE )
Measures of Uncertainty
Extract the coefficient, standard error, confidence interval, and coefficient of variation from any descriptive statistics function result, overall and by groups:
this_result <- svymean( ~ h12ahous , hrs_design , na.rm = TRUE )
coef( this_result )
SE( this_result )
confint( this_result )
cv( this_result )
grouped_result <-
svyby(
~ h12ahous ,
~ marital_stat_1996 ,
hrs_design ,
svymean ,
na.rm = TRUE
)
coef( grouped_result )
SE( grouped_result )
confint( grouped_result )
cv( grouped_result )
Calculate the degrees of freedom of any survey design object:
degf( hrs_design )
Calculate the complex sample survey-adjusted variance of any statistic:
svyvar( ~ h12ahous , hrs_design , na.rm = TRUE )
Include the complex sample design effect in the result for a specific statistic:
# SRS without replacement
svymean( ~ h12ahous , hrs_design , na.rm = TRUE , deff = TRUE )
# SRS with replacement
svymean( ~ h12ahous , hrs_design , na.rm = TRUE , deff = "replace" )
Compute confidence intervals for proportions using methods that may be more accurate near 0 and 1. See ?svyciprop
for alternatives:
svyciprop( ~ working_in_2014 , hrs_design ,
method = "likelihood" , na.rm = TRUE )
Regression Models and Tests of Association
Perform a design-based t-test:
svyttest( h12ahous ~ working_in_2014 , hrs_design )
Perform a chi-squared test of association for survey data:
svychisq(
~ working_in_2014 + marital_stat_2014 ,
hrs_design
)
Perform a survey-weighted generalized linear model:
glm_result <-
svyglm(
h12ahous ~ working_in_2014 + marital_stat_2014 ,
hrs_design
)
summary( glm_result )
Analysis Examples with srvyr
The R srvyr
library calculates summary statistics from survey data, such as the mean, total or quantile using dplyr-like syntax. srvyr allows for the use of many verbs, such as summarize
, group_by
, and mutate
, the convenience of pipe-able functions, the tidyverse
style of non-standard evaluation and more consistent return types than the survey
package. This vignette details the available features. As a starting point for HRS users, this code replicates previously-presented examples:
library(srvyr)
hrs_srvyr_design <- as_survey( hrs_design )
Calculate the mean (average) of a linear variable, overall and by groups:
hrs_srvyr_design %>%
summarize( mean = survey_mean( h12ahous , na.rm = TRUE ) )
hrs_srvyr_design %>%
group_by( marital_stat_1996 ) %>%
summarize( mean = survey_mean( h12ahous , na.rm = TRUE ) )