Stratified Sampling: Software
(Stata, R)

PUBHBIO 7225 Lecture 8

Outline

Topics

• Stata for Stratified Samples
• R for Stratified Samples

Activities

• 8.1 Stata/R for Stratified Samples

Assignments

• Peer Evaluation of Problem Set 2 due Tuesday 9/23/25 11:59pm via Carmen
• Quiz 2 due Thursday 9/25/2025 11:59pm via Carmen
• Group Progress Report due Thursday 9/25/2025 11:59pm via Carmen (only one group member needs to upload this)

Stata Commands for Stratified Samples

Select a stratified random sample of \(n_h=50\) in each stratum:

sort STRATVAR IDVAR     # STRATVAR = variable defining strata, IDVAR = record ID
set seed 217 
by STRATVAR: sample 50, count

Select a stratified random sample by taking 50% of each stratum:

sort STRATVAR IDVAR 
set seed 531 
by STRATVAR: sample 50 

Tell Stata about the sample design for a stratified sample:

svyset [pweight=WEIGHTVAR], fpc(N_h) strata(STRATVAR) 
# WEIGHTVAR = variable containing sampling weights
# N_h = variable that contains finite population sizes in each stratum

With the survey design set, for estimation use same commands as for SRS (e.g., svy:)

R Commands for Stratified Samples

Select a stratified random sample of \(n_h=50\) in each stratum:

set.seed(217)
DF <- POPDATA %>% group_by(STRATVAR) %>% slice_sample(n = 50)
# STRATVAR = variable defining strata

Select a stratified random sample by taking 50% of each stratum:

set.seed(217)
DF <- POPDATA %>% group_by(STRATVAR) %>% slice_sample(prop = 0.5)

Tell R about the sample design for a stratified sample:

DESIGN.OBJECT <- svydesign(id = ~1, weights = ~WEIGHTVAR, fpc = ~N_h, 
                           strata = ~STRATVAR, data=DF)
# WEIGHTVAR = variable containing sampling weights
# N_h = variable that contains finite population sizes in each stratum

With the survey design object created, for estimation use same commands as for SRS (e.g., svymean())

Activity 8.1

Stata/R for Stratified Samples