Package {svymargins}

Title: Predictive Margins for Survey Analyses
Version: 0.1.0
Author: Tommi Härkänen [aut, cre]
Maintainer: Tommi Härkänen <tommi.harkanen@thl.fi>
Depends: R (≥ 4.2.0)
Description: Predictive margins like in the 'Stata' procedure 'margins'. This package is based on the survey::svypredmean() function. Supported regression analyses are based on survey::svyglm() and svyVGAM::svy_vglm() functions (for multinomial logistic regression models).
License: GPL (≥ 3)
Encoding: UTF-8
Imports: plyr, dplyr, stringr, survey, tidyselect, VGAM
Suggests: knitr, rmarkdown, svyVGAM
VignetteBuilder: knitr
Config/roxygen2/version: 8.0.0
NeedsCompilation: no
Packaged: 2026-07-01 11:27:55 UTC; thah
Repository: CRAN
Date/Publication: 2026-07-06 14:30:02 UTC

Helper function to convert symmetric probability scale confidence intervals into asymmetric

Description

Helper function to convert symmetric probability scale confidence intervals into asymmetric

Usage

asymm_ci(x, ci.level = 0.95)

Arguments

x

Matrix with two columns as returned by confint function.

ci.level

Default 0.95 for 95% confidence intervals.

The transformation is based on the logit transformation and the delta method.

Value

Matrix with two columns as returned by confint function.

Examples

library(survey)
library(dplyr)
library(svyVGAM)
n <- 1000
# Generate data:
set.seed(1234)
d <- data.frame(sex=factor(sample(c("M", "F"), n, replace=TRUE)),
                education=factor(sample(c("low", "middle", "high"), n, replace=TRUE)))
d <- d |>
  mutate(age=runif(n, 0, 40) + as.numeric(education) * 20,
         pr1=1,
         pr2=exp(-3 + 0.5 * as.numeric(education) + 0.05 * age),
         pr3=exp(1 + -0.5 * as.numeric(education) + 0.02 * age)) |> 
  rowwise() |>
  mutate(y=which(rmultinom(1, 1, c(pr1, pr2, pr3))[,1] == 1)) |> 
  ungroup() |>
  mutate(y=factor(y, levels=1:3, labels=LETTERS[1:3]))
# Create survey design:
my.svy <- svydesign(~ 1, weights=~ 1, data=d)
# Run regression analysis:
res <- svy_vglm(y ~ education + age + sex, family=multinomial(refLevel=1), design=my.svy)
# Define margins as a named list:
target.l <- list(null=list(),
                 educ=list("education"),
                 age=list(age=seq(40,70,10)),
                 educ_age=list("education", age=seq(40,70,10)))
# Calculate predictive margins:
marg <- svymargins(res, groupfactor=target.l, y.lev=2)
confint(marg)
confint(marg) |> asymm_ci()

Specify constraints for selected covariates for predicted margins

Description

Specify constraints for selected covariates for predicted margins

Usage

scenarios2group(scenarios, dat)

Arguments

scenarios

A (named) list of scenarios based on factor and/or continuous covariates.

dat

A dataframe containing the covariates.

Details

All combinations of the specified covariates are returned as a dataframe. If the list contains more than one element, then the dataframes based on the elements are returned, and binded together. if scenarios is a named list, then the returned dataframe has column group_id containing the list names.

Value

A dataframe containing the combinations of the specified covariates.

Examples

n <- 1000
d <- data.frame(sex=factor(sample(c("M", "F"), n, replace=TRUE)),
                age=runif(20, 100, n),
                education=factor(sample(c("low", "middle", "high"), n, replace=TRUE)),
                y=rnorm(n))
target.l <- list(null=list(), educ=list("education"), age=list(age=seq(40,70,10)),
                 educ_age=list("education", age=seq(40,70,10)))
scenarios2group(target.l, d)

Generic function for svymargins

Description

Generic function for svymargins

Usage

svymargins(adjustmodel, ...)

Arguments

adjustmodel

A regression analysis result object.

...

Other arguments

See other methods for different regression analyses.

Value

As svypredmeans, an object of class svystat with the predictive marginal means and their covariance matrix. Additional attribute: groups, which contains dataframe with the groupfactor variable values, and the estimated margins and standard errors.

The default method for svymargins

Description

The default method for svymargins

Usage

## Default S3 method:
svymargins(adjustmodel, ...)

Arguments

adjustmodel

A regression analysis result object.

...

Other arguments

Executed for regression analyses, which have not been implemented.

Value

An error message.

See Also

svymargins.svyglm, svymargins.svy_vglm

Predictive margins for multinomial logistic regression model

Description

Predictive margins for multinomial logistic regression model

Usage

## S3 method for class 'svy_vglm'
svymargins(adjustmodel, ..., groupfactor, y.lev = NULL, subs = NULL)

Arguments

adjustmodel

Result of regression analysis object from the survey::svyglm function.

...

Currently unused.

groupfactor

Specification of the margins: Character vector of factor variable names or a list. See details and examples.

y.lev

Output category (name or number of level) for which margins are calculated.

subs

Character string specifying a subset, e.g. "age < 50 & year == 2015".

Details

groupfactor is a list specifying several margins specifications. The list can contain one or more lists. A list can contain the names of factor covariates (as character strings) and/or continuous covariates as named vectors containing the desired values at which to calculate the margins.

Value

As svypredmeans, an object of class svystat with the predictive marginal means and their covariance matrix. Additional attribute: groups, which contains dataframe with the groupfactor variable values, and the estimated margins and standard errors. The result of svymargins is named based on the groupfactor, but if there are duplicates, then the result is unnamed.

Examples

library(survey)
library(dplyr)
library(svyVGAM)
n <- 1000
# Generate data:
set.seed(1234)
d <- data.frame(sex=factor(sample(c("M", "F"), n, replace=TRUE)),
                education=factor(sample(c("low", "middle", "high"), n, replace=TRUE)))
d <- d |>
  mutate(age=runif(n, 0, 40) + as.numeric(education) * 20,
         pr1=1,
         pr2=exp(-3 + 0.5 * as.numeric(education) + 0.05 * age),
         pr3=exp(1 + -0.5 * as.numeric(education) + 0.02 * age)) |> 
  rowwise() |>
  mutate(y=which(rmultinom(1, 1, c(pr1, pr2, pr3))[,1] == 1)) |> 
  ungroup() |>
  mutate(y=factor(y, levels=1:3, labels=LETTERS[1:3]))
# Create survey design:
my.svy <- svydesign(~ 1, weights=~ 1, data=d)
# Run regression analysis:
res <- svy_vglm(y ~ education + age + sex, family=multinomial(refLevel=1), design=my.svy)
# Define margins as a named list:
target.l <- list(null=list(),
                 educ=list("education"),
                 age=list(age=seq(40,70,10)),
                 educ_age=list("education", age=seq(40,70,10)))
# Calculate predictive margins:
svymargins(res, groupfactor=target.l, y.lev=2)
# Get the output table containing the covariate information from "groups" attribute:
attr(svymargins(res, groupfactor=target.l, y.lev="B"), "groups")

Mimics the margins procedure of Stata for GLM estimated with survey::svyglm

Description

Based on the survey::svypredmeans function, but is more flexible. No restrictions with interactions. Implements also subsets with the subs argument.

Usage

## S3 method for class 'svyglm'
svymargins(adjustmodel, ..., groupfactor, y.lev = NULL, subs = NULL)

Arguments

adjustmodel

Result of regression analysis object from the survey::svyglm function.

...

Currently unused.

groupfactor

Specification of the margins: Character vector of factor variable names or a list. See details and examples.

y.lev

Not currently used in svyglm models.

subs

Character string specifying a subset, e.g. "age < 50 & year == 2015".

Details

groupfactor is a list specifying several margins specifications. The list can contain one or more lists. A list can contain the names of factor covariates (as character strings) and/or continuous covariates as named vectors containing the desired values at which to calculate the margins.

Value

As svypredmeans, an object of class svystat with the predictive marginal means and their covariance matrix. Additional attribute: groups, which contains dataframe with the groupfactor variable values, and the estimated margins and standard errors. The result of svymargins is named based on the groupfactor, but if there are duplicates, then the result is unnamed.

Examples

library(survey)
library(dplyr)
n <- 1000
# Generate data:
set.seed(1234)
d <- data.frame(sex=factor(sample(c("M", "F"), n, replace=TRUE)),
                education=factor(sample(c("low", "middle", "high"), n, replace=TRUE)))
d <- d |> mutate(age=runif(n, 0, 40) + as.numeric(education) * 20,
                 y=rnorm(n, sd=5) + as.numeric(education) + 0.05 * age)
# Create survey design:
my.svy <- svydesign(~ 1, weights=~ 1, data=d)
# Run regression analysis:
res <- svyglm(y ~ education + age + sex, design=my.svy)
# Define margins as a named list:
target.l <- list(null=list(),
                 educ=list("education"),
                 age=list(age=seq(40,70,10)),
                 educ_age=list("education", age=seq(40,70,10)))
# Calculate predictive margins:
svymargins(res, groupfactor=target.l)
# Get the output table containing the covariate information from "groups" attribute:
attr(svymargins(res, groupfactor=target.l), "groups")