Complex survey analysis

Start with the expanded example for each command. Open the variants for options and alternative R approaches. Read the qualifications before adapting the code.

estat

Coefficient of variation of a survey mean

Stata

estat cv

R survey

survey::cv(survey::svymean(~y, des))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Relative standard error is unstable when the estimated mean is near zero.

Stata reference · R reference

1 more examples and alternatives

Design effect of a survey mean

Stata

estat effects

R survey

survey::deff(survey::svymean(~y, des, deff = TRUE))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design effects depend on the SRS comparison, finite-population correction and weight scaling.

Stata reference · R reference

svy

Survey mean with bootstrap replicate variance

Stata

svy bootstrap: mean y

R survey

{ reps <- survey::as.svrepdesign(des, type = "bootstrap"); survey::svymean(~y, reps) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Point means agree with sampling weights; replication variance depends on the chosen scheme.

Stata reference · R reference

30 more examples and alternatives

Survey mean with BRR replicate variance

Stata

svy brr: mean y

R survey

{ reps <- survey::as.svrepdesign(des, type = "BRR"); survey::svymean(~y, reps) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Point means agree with sampling weights; replication variance depends on the chosen scheme.

Stata reference · R reference

Survey mean with JKn replicate variance

Stata

svy jackknife: mean y

R survey

{ reps <- survey::as.svrepdesign(des, type = "JKn"); survey::svymean(~y, reps) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Point means agree with sampling weights; replication variance depends on the chosen scheme.

Stata reference · R reference

Survey subpopulation mean

Stata

svy, subpop(if group==1): mean y

R survey

survey::svymean(~y, subset(des, group == "1"))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Subset the design, preserving the sampling structure, rather than rebuilding a design after filtering.

Stata reference · R reference

Survey gamma log-link model

Stata

svy: glm y x, family(gamma) link(log)

R survey

survey::svyglm(y ~ x, des, family = stats::Gamma(link = "log"))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Quasi families retain the mean/variance structure without non-integer count warnings; align design degrees of freedom.

Stata reference · R reference

Survey logistic regression

Stata

svy: logit yes x

R survey

survey::svyglm(yes ~ x, des, family = stats::quasibinomial())

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Quasi families retain the mean/variance structure without non-integer count warnings; align design degrees of freedom.

Stata reference · R reference

Survey-weighted mean

Stata

svy: mean y

R survey

survey::svymean(~y, des)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Joint means and covariance matrix

Stata

svy: mean y x

R survey

survey::svymean(~y + x, des)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Domain-specific means

Stata

svy: mean y, over(group)

R survey

survey::svyby(~y, ~group, des, survey::svymean)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Survey Poisson mean model

Stata

svy: poisson y x

R survey

survey::svyglm(y ~ x, des, family = stats::quasipoisson())

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Quasi families retain the mean/variance structure without non-integer count warnings; align design degrees of freedom.

Stata reference · R reference

Survey probit regression

Stata

svy: probit yes x

R survey

survey::svyglm(yes ~ x, des, family = stats::quasibinomial(link = "probit"))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Quasi families retain the mean/variance structure without non-integer count warnings; align design degrees of freedom.

Stata reference · R reference

Survey proportion interval using asin

Stata

svy: proportion yes

R survey

survey::svyciprop(~yes, des, method = "asin", level = .95)

Fixed workflow; adapt the named data columns explicitly. Different interval constructions are intentional alternatives; match Stata’s selected transformation.

Stata reference · R reference

Survey proportion interval using beta

Stata

svy: proportion yes

R survey

survey::svyciprop(~yes, des, method = "beta", level = .95)

Fixed workflow; adapt the named data columns explicitly. Different interval constructions are intentional alternatives; match Stata’s selected transformation.

Stata reference · R reference

Survey proportion interval using likelihood

Stata

svy: proportion yes

R survey

survey::svyciprop(~yes, des, method = "likelihood", level = .95)

Fixed workflow; adapt the named data columns explicitly. Different interval constructions are intentional alternatives; match Stata’s selected transformation.

Stata reference · R reference

Survey proportion interval using logit

Stata

svy: proportion yes

R survey

survey::svyciprop(~yes, des, method = "logit", level = .95)

Fixed workflow; adapt the named data columns explicitly. Different interval constructions are intentional alternatives; match Stata’s selected transformation.

Stata reference · R reference

Survey proportion interval using mean

Stata

svy: proportion yes

R survey

survey::svyciprop(~yes, des, method = "mean", level = .95)

Fixed workflow; adapt the named data columns explicitly. Different interval constructions are intentional alternatives; match Stata’s selected transformation.

Stata reference · R reference

Survey proportion interval using xlogit

Stata

svy: proportion yes

R survey

survey::svyciprop(~yes, des, method = "xlogit", level = .95)

Fixed workflow; adapt the named data columns explicitly. Different interval constructions are intentional alternatives; match Stata’s selected transformation.

Stata reference · R reference

Survey-weighted category proportions

Stata

svy: proportion yes

R survey

survey::svymean(~factor(yes), des)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Domain-specific proportions

Stata

svy: proportion yes, over(group)

R survey

survey::svyby(~factor(yes), ~group, des, survey::svymean)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Survey ratio of totals

Stata

svy: ratio y/x

R survey

survey::svyratio(~y, ~x, des)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Survey linear regression

Stata

svy: regress y x

R survey

survey::svyglm(y ~ x, des, family = stats::gaussian())

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Quasi families retain the mean/variance structure without non-integer count warnings; align design degrees of freedom.

Stata reference · R reference

Weighted contingency-table counts

Stata

svy: tabulate group yes

R survey

prop.table(survey::svytable(~group + yes, des))

Differs from Stata Returns weighted cell proportions, the default display of svy: tabulate twoway. This table alone does not provide Stata design-based standard errors, confidence intervals or adjusted association tests.

Stata reference · R reference

Survey column proportions

Stata

svy: tabulate group yes, column

R survey

prop.table(survey::svytable(~group + yes, des), margin = 2)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Design-based categorical association: Chisq

Stata

svy: tabulate group yes, pearson

R survey

survey::svychisq(~group + yes, des, statistic = "Chisq")

Fixed workflow; adapt the named data columns explicitly. Rao-Scott and Wald choices have distinct reference distributions; choose the intended test explicitly.

Stata reference · R reference

Design-based categorical association: F

Stata

svy: tabulate group yes, pearson

R survey

survey::svychisq(~group + yes, des, statistic = "F")

Fixed workflow; adapt the named data columns explicitly. Rao-Scott and Wald choices have distinct reference distributions; choose the intended test explicitly.

Stata reference · R reference

Design-based categorical association: Wald

Stata

svy: tabulate group yes, pearson

R survey

survey::svychisq(~group + yes, des, statistic = "Wald")

Fixed workflow; adapt the named data columns explicitly. Rao-Scott and Wald choices have distinct reference distributions; choose the intended test explicitly.

Stata reference · R reference

Design-based categorical association: adjWald

Stata

svy: tabulate group yes, pearson

R survey

survey::svychisq(~group + yes, des, statistic = "adjWald")

Fixed workflow; adapt the named data columns explicitly. Rao-Scott and Wald choices have distinct reference distributions; choose the intended test explicitly.

Stata reference · R reference

Survey row proportions

Stata

svy: tabulate group yes, row

R survey

prop.table(survey::svytable(~group + yes, des), margin = 1)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Survey-weighted population total

Stata

svy: total y

R survey

survey::svytotal(~y, des)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Joint population totals

Stata

svy: total y x

R survey

survey::svytotal(~y + x, des)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

Domain-specific population totals

Stata

svy: total y, over(group)

R survey

survey::svyby(~y, ~group, des, survey::svytotal)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Design-based variance uses the full sampling design; population totals require correctly scaled weights.

Stata reference · R reference

svydescribe

Survey design degrees of freedom and cluster counts

Stata

svydescribe

R survey

list(degrees_of_freedom = survey::degf(des), strata = table(des$strata[, 1]), clusters = length(unique(des$cluster[, 1])))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Counts observed records by stratum; PSU count is distinct from observation count.

Stata reference · R reference

svyset

Independent unequal-probability sample

Stata

svyset [pw=w]

R survey

survey::svydesign(ids = ~1, weights = ~w, data = data)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Survey objects carry design metadata; do not substitute ordinary regression weights.

Stata reference · R reference

8 more examples and alternatives

Stratified cluster survey design

Stata

svyset psu [pw=w], strata(stratum)

R survey

survey::svydesign(ids = ~psu, strata = ~stratum, weights = ~w, data = data)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Survey objects carry design metadata; do not substitute ordinary regression weights.

Stata reference · R reference

Finite-population correction in a clustered design

Stata

svyset psu [pw=w], strata(stratum) fpc(N)

R survey

survey::svydesign(ids = ~psu, strata = ~stratum, weights = ~w, fpc = ~N, data = data)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Survey objects carry design metadata; do not substitute ordinary regression weights.

Stata reference · R reference

Poststratify survey weights to population category totals

Stata

svyset psu [pw=w], strata(stratum) poststrata(group) postweight(pop)

R survey

survey::postStratify(des, ~group, population = population)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Population categories must cover the sample; weighting adjustments are part of the variance design.

Stata reference · R reference

Rake sample weights to known category margins

Stata

svyset psu [pw=w], strata(stratum) rake(ibn.group, totals(20 30, copy) noconstant)

R survey

survey::rake(des, sample.margins = list(~group), population.margins = list(population))

Fixed workflow; adapt the named data columns explicitly. Survey calibration workflow; Stata’s exact calibration syntax depends on svyset options/version.

Stata reference · R reference

Survey bootstrap replicate-weight design

Stata

svyset psu [pw=w], strata(stratum) vce(bootstrap)

R survey

survey::as.svrepdesign(des, type = "bootstrap", replicates = 50)

Fixed workflow; adapt the named data columns explicitly. Illustrative replication count; specific survey bootstrap schemes and scales can differ.

Stata reference · R reference

Balanced repeated replication for paired-PSU strata

Stata

svyset psu [pw=w], strata(stratum) vce(brr)

R survey

survey::as.svrepdesign(des, type = "BRR")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. BRR requires exactly two PSUs per stratum unless a deliberate design adjustment is made.

Stata reference · R reference

Fay balanced repeated replication

Stata

svyset psu [pw=w], strata(stratum) vce(brr) fay(.3)

R survey

survey::as.svrepdesign(des, type = "Fay", fay.rho = .3)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match Fay’s perturbation parameter and replicate scaling conventions.

Stata reference · R reference

Stratified delete-one jackknife replicates

Stata

svyset psu [pw=w], vce(jackknife)

R survey

survey::as.svrepdesign(des, type = "JKn")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Survey objects carry design metadata; do not substitute ordinary regression weights.

Stata reference · R reference