Group and ungroup a survey design object

group_by() stores grouping columns on the survey object for use in grouped operations like mutate(). ungroup() removes the grouping. group_vars() returns the current grouping column names.

Unlike dplyr, groups are not attached to the underlying data frame — they are stored on the survey object itself and applied when needed by verbs that support grouping.

Usage

# S3 method for class 'survey_base'
group_by(.data, ..., .add = FALSE, .drop = dplyr::group_by_drop_default(.data))

# S3 method for class 'survey_base'
ungroup(x, ...)

group_by(.data, ..., .add = FALSE, .drop = group_by_drop_default(.data))

Arguments

.data: A survey_base object.
...: <data-masking> For group_by(): columns to group by. Computed expressions (e.g., cut(ridageyr, breaks = c(0, 18, 65, Inf))) are supported. For ungroup(): columns to remove from the current grouping. Omit to remove all groups.
.add: When FALSE (default), replaces existing groups. Use .add = TRUE to add to the current grouping instead.
.drop: Accepted for compatibility with the dplyr interface; has no effect on survey design objects.
x: A survey_base object (for ungroup() and group_vars()).

Value

An object of the same type as the input with the following properties:

Rows, columns, and survey design attributes are unchanged.
For group_by(): grouping columns are set or updated; rowwise keys are cleared.
For ungroup(): all or specified grouping columns are removed; rowwise keys are cleared on full ungroup only.
For group_vars(): a character vector of current grouping column names.

Details

Grouped operations

After calling group_by(), mutate() computes within groups. Future estimation functions will also use grouping to perform stratified analysis.

Adding to existing groups

By default, group_by() replaces existing groups. Use .add = TRUE to append to the current grouping instead.

Rowwise mode and group_by()

group_by(.add = FALSE) (the default) exits rowwise mode — it clears @variables$rowwise and @variables$rowwise_id_cols. group_by(.add = TRUE) when the design is rowwise promotes the rowwise id columns to @groups, appends the new groups, then clears rowwise mode — mirroring dplyr's behaviour exactly.

Partial ungroup

ungroup() with no arguments removes all groups and exits rowwise mode. With column arguments, it removes only the specified columns from the grouping — rowwise mode is not affected.

Examples

library(surveytidy)
library(surveycore)
library(dplyr)
d <- as_survey(pew_npors_2025, weights = weight, strata = stratum)

# Group by a column
group_by(d, gender)
#> 
#> ── Survey Design ───────────────────────────────────────────────────────────────
#> <survey_taylor> (Taylor series linearization)
#> Sample size: 5022
#> Groups: gender
#> 
#> # A tibble: 5,022 × 65
#>    respid  mode language languageinitial stratum interview_start interview_end
#>     <dbl> <dbl>    <dbl>           <dbl>   <dbl> <date>          <date>       
#>  1   1470     2        1              NA      10 2025-05-27      2025-05-27   
#>  2   2374     2        1              NA       7 2025-05-01      2025-05-01   
#>  3   1177     3        1              10       5 2025-03-04      2025-03-04   
#>  4  15459     2        1              NA      10 2025-05-05      2025-05-05   
#>  5   9849     1        1               9       9 2025-02-22      2025-02-22   
#>  6   8178     3        1               9      10 2025-03-10      2025-03-10   
#>  7   3682     1        1               9       4 2025-02-27      2025-02-27   
#>  8   6999     2        1              NA      10 2025-05-12      2025-05-12   
#>  9   9945     2        1              NA      10 2025-05-09      2025-05-09   
#> 10   1901     1        1               9      10 2025-03-01      2025-03-01   
#> # ℹ 5,012 more rows
#> # ℹ 58 more variables: econ1mod <dbl>, econ1bmod <dbl>, comtype2 <dbl>,
#> #   unity <dbl>, crimesafe <dbl>, govprotct <dbl>, moregunimpact <dbl>,
#> #   fin_sit <dbl>, vet1 <dbl>, vol12_cps <dbl>, eminuse <dbl>, intmob <dbl>,
#> #   intfreq <dbl>, intfreq_collapsed <dbl>, home4nw2 <dbl>, bbhome <dbl>,
#> #   smuse_fb <dbl>, smuse_yt <dbl>, smuse_x <dbl>, smuse_ig <dbl>,
#> #   smuse_sc <dbl>, smuse_wa <dbl>, smuse_tt <dbl>, smuse_rd <dbl>, …

# Grouped mutate — within-group mean centring
d |>
  group_by(gender) |>
  mutate(econ_centred = econ1mod - mean(econ1mod, na.rm = TRUE))
#> 
#> ── Survey Design ───────────────────────────────────────────────────────────────
#> <survey_taylor> (Taylor series linearization)
#> Sample size: 5022
#> Groups: gender
#> 
#> # A tibble: 5,022 × 66
#>    respid  mode language languageinitial stratum interview_start interview_end
#>     <dbl> <dbl>    <dbl>           <dbl>   <dbl> <date>          <date>       
#>  1   1470     2        1              NA      10 2025-05-27      2025-05-27   
#>  2   2374     2        1              NA       7 2025-05-01      2025-05-01   
#>  3   1177     3        1              10       5 2025-03-04      2025-03-04   
#>  4  15459     2        1              NA      10 2025-05-05      2025-05-05   
#>  5   9849     1        1               9       9 2025-02-22      2025-02-22   
#>  6   8178     3        1               9      10 2025-03-10      2025-03-10   
#>  7   3682     1        1               9       4 2025-02-27      2025-02-27   
#>  8   6999     2        1              NA      10 2025-05-12      2025-05-12   
#>  9   9945     2        1              NA      10 2025-05-09      2025-05-09   
#> 10   1901     1        1               9      10 2025-03-01      2025-03-01   
#> # ℹ 5,012 more rows
#> # ℹ 59 more variables: econ1mod <dbl>, econ1bmod <dbl>, comtype2 <dbl>,
#> #   unity <dbl>, crimesafe <dbl>, govprotct <dbl>, moregunimpact <dbl>,
#> #   fin_sit <dbl>, vet1 <dbl>, vol12_cps <dbl>, eminuse <dbl>, intmob <dbl>,
#> #   intfreq <dbl>, intfreq_collapsed <dbl>, home4nw2 <dbl>, bbhome <dbl>,
#> #   smuse_fb <dbl>, smuse_yt <dbl>, smuse_x <dbl>, smuse_ig <dbl>,
#> #   smuse_sc <dbl>, smuse_wa <dbl>, smuse_tt <dbl>, smuse_rd <dbl>, …

# Add a second grouping variable with .add = TRUE
d |>
  group_by(gender) |>
  group_by(cregion, .add = TRUE)
#> 
#> ── Survey Design ───────────────────────────────────────────────────────────────
#> <survey_taylor> (Taylor series linearization)
#> Sample size: 5022
#> Groups: gender and cregion
#> 
#> # A tibble: 5,022 × 65
#>    respid  mode language languageinitial stratum interview_start interview_end
#>     <dbl> <dbl>    <dbl>           <dbl>   <dbl> <date>          <date>       
#>  1   1470     2        1              NA      10 2025-05-27      2025-05-27   
#>  2   2374     2        1              NA       7 2025-05-01      2025-05-01   
#>  3   1177     3        1              10       5 2025-03-04      2025-03-04   
#>  4  15459     2        1              NA      10 2025-05-05      2025-05-05   
#>  5   9849     1        1               9       9 2025-02-22      2025-02-22   
#>  6   8178     3        1               9      10 2025-03-10      2025-03-10   
#>  7   3682     1        1               9       4 2025-02-27      2025-02-27   
#>  8   6999     2        1              NA      10 2025-05-12      2025-05-12   
#>  9   9945     2        1              NA      10 2025-05-09      2025-05-09   
#> 10   1901     1        1               9      10 2025-03-01      2025-03-01   
#> # ℹ 5,012 more rows
#> # ℹ 58 more variables: econ1mod <dbl>, econ1bmod <dbl>, comtype2 <dbl>,
#> #   unity <dbl>, crimesafe <dbl>, govprotct <dbl>, moregunimpact <dbl>,
#> #   fin_sit <dbl>, vet1 <dbl>, vol12_cps <dbl>, eminuse <dbl>, intmob <dbl>,
#> #   intfreq <dbl>, intfreq_collapsed <dbl>, home4nw2 <dbl>, bbhome <dbl>,
#> #   smuse_fb <dbl>, smuse_yt <dbl>, smuse_x <dbl>, smuse_ig <dbl>,
#> #   smuse_sc <dbl>, smuse_wa <dbl>, smuse_tt <dbl>, smuse_rd <dbl>, …

# Remove all groups
d |> group_by(gender) |> ungroup()
#> 
#> ── Survey Design ───────────────────────────────────────────────────────────────
#> <survey_taylor> (Taylor series linearization)
#> Sample size: 5022
#> 
#> # A tibble: 5,022 × 65
#>    respid  mode language languageinitial stratum interview_start interview_end
#>     <dbl> <dbl>    <dbl>           <dbl>   <dbl> <date>          <date>       
#>  1   1470     2        1              NA      10 2025-05-27      2025-05-27   
#>  2   2374     2        1              NA       7 2025-05-01      2025-05-01   
#>  3   1177     3        1              10       5 2025-03-04      2025-03-04   
#>  4  15459     2        1              NA      10 2025-05-05      2025-05-05   
#>  5   9849     1        1               9       9 2025-02-22      2025-02-22   
#>  6   8178     3        1               9      10 2025-03-10      2025-03-10   
#>  7   3682     1        1               9       4 2025-02-27      2025-02-27   
#>  8   6999     2        1              NA      10 2025-05-12      2025-05-12   
#>  9   9945     2        1              NA      10 2025-05-09      2025-05-09   
#> 10   1901     1        1               9      10 2025-03-01      2025-03-01   
#> # ℹ 5,012 more rows
#> # ℹ 58 more variables: econ1mod <dbl>, econ1bmod <dbl>, comtype2 <dbl>,
#> #   unity <dbl>, crimesafe <dbl>, govprotct <dbl>, moregunimpact <dbl>,
#> #   fin_sit <dbl>, vet1 <dbl>, vol12_cps <dbl>, eminuse <dbl>, intmob <dbl>,
#> #   intfreq <dbl>, intfreq_collapsed <dbl>, home4nw2 <dbl>, bbhome <dbl>,
#> #   smuse_fb <dbl>, smuse_yt <dbl>, smuse_x <dbl>, smuse_ig <dbl>,
#> #   smuse_sc <dbl>, smuse_wa <dbl>, smuse_tt <dbl>, smuse_rd <dbl>, …

# Partial ungroup — remove only gender, keep cregion
d |>
  group_by(gender, cregion) |>
  ungroup(gender)
#> 
#> ── Survey Design ───────────────────────────────────────────────────────────────
#> <survey_taylor> (Taylor series linearization)
#> Sample size: 5022
#> Groups: cregion
#> 
#> # A tibble: 5,022 × 65
#>    respid  mode language languageinitial stratum interview_start interview_end
#>     <dbl> <dbl>    <dbl>           <dbl>   <dbl> <date>          <date>       
#>  1   1470     2        1              NA      10 2025-05-27      2025-05-27   
#>  2   2374     2        1              NA       7 2025-05-01      2025-05-01   
#>  3   1177     3        1              10       5 2025-03-04      2025-03-04   
#>  4  15459     2        1              NA      10 2025-05-05      2025-05-05   
#>  5   9849     1        1               9       9 2025-02-22      2025-02-22   
#>  6   8178     3        1               9      10 2025-03-10      2025-03-10   
#>  7   3682     1        1               9       4 2025-02-27      2025-02-27   
#>  8   6999     2        1              NA      10 2025-05-12      2025-05-12   
#>  9   9945     2        1              NA      10 2025-05-09      2025-05-09   
#> 10   1901     1        1               9      10 2025-03-01      2025-03-01   
#> # ℹ 5,012 more rows
#> # ℹ 58 more variables: econ1mod <dbl>, econ1bmod <dbl>, comtype2 <dbl>,
#> #   unity <dbl>, crimesafe <dbl>, govprotct <dbl>, moregunimpact <dbl>,
#> #   fin_sit <dbl>, vet1 <dbl>, vol12_cps <dbl>, eminuse <dbl>, intmob <dbl>,
#> #   intfreq <dbl>, intfreq_collapsed <dbl>, home4nw2 <dbl>, bbhome <dbl>,
#> #   smuse_fb <dbl>, smuse_yt <dbl>, smuse_x <dbl>, smuse_ig <dbl>,
#> #   smuse_sc <dbl>, smuse_wa <dbl>, smuse_tt <dbl>, smuse_rd <dbl>, …

# Get current grouping column names
d |> group_by(gender, cregion) |> group_vars()
#> [1] "gender"  "cregion"