gtsummary 中调查回复的平均分数

Question

我正在尝试使用

tbl_svysummary

函数按 gtsummary 中的分类变量生成加权平均调查分数表。

如何使用调查权重、使用

tbl_svysummary

功能来执行此操作？我已经浏览了两个与 Daniel Sjoberg 提供的解决方案很接近的示例，但我只是不太明白它们发生了什么。我可以通过调换

by

中的

include

和

tbl_svysummary

参数来做到这一点，但随后我遇到了一个全新的问题，即表格方向错误。

这是一些示例数据：

dat <- structure(list(uuid = c("p41112019021430", "p41222013024584", 
"p41212017017560", "p41212017011700", "p41212019022003", "p41212019133026", 
"p41212017014434", "p41112019023063", "p41212019077561", "p41212017050030"
), age_cat = structure(c(3L, 1L, 4L, 2L, 2L, 3L, 3L, 2L, 1L, 
4L), levels = c("18-24", "25-44", "45-64", "65-74", "75+"), class = "factor"), 
    cvh_score = c(6, 4, 1, 3, 0, 2, 3, 2, 6, 1), weights = c(p41112019021430 = 0.360602284454939, 
    p41222013024584 = 5.00004172246093, p41212017017560 = 0.276025143197602, 
    p41212017011700 = 1.55086389757734, p41212019022003 = 2.20669366738008, 
    p41212019133026 = 0.878664071962474, p41212017014434 = 1.15252329666968, 
    p41112019023063 = 1.51638372307208, p41212019077561 = 2.1408232841115, 
    p41212017050030 = 0.282529671403006)), row.names = c(NA, 
-10L), class = "data.frame")

dat_svy <- dat |> 
  srvyr::as_survey_design(
    ids = uuid, 
    weights = weights
  )

我可以使用

tbl_continuous

函数在没有权重的情况下实现我想要的目标，如下所示：

tbl_01 <- gtsummary::tbl_continuous(
  dat, 
  variable = cvh_score, 
  include = c(age_cat), 
  statistic = list(
    everything() ~ "{mean} ({sd})"
  )
  ) 

tbl_01

这给了我以下内容：

输出示例

如何用重量做到这一点？

Answer 1

这是我想出的解决方案。我确信有人可以做得更干净。

weighted_mean <- dat_svy |>
  filter(!is.na(age_cat)) |>
  group_by(age_cat) |>
  summarize(mean = survey_mean(cvh_score), sd = survey_sd(cvh_score)) |>
  transmute(
    row_type = "level",
    label = age_cat,
    stat_3 = str_glue("{round(mean, 2)} ({round(sd, 2)})")
  ) |>
  add_row(
    row_type = "label",
    label = "Age",
    stat_3 = NA,
    .before = 1L
  )

table <- tbl_svysummary(
  dat_svy, 
  include = c("age_cat"), 
  label = list(
    age_cat ~ "Age" 
  ), 
  statistic = list(
    all_categorical() ~ "{n}"
  ),
  percent = "row", 
  missing = "no"
  ) |> 
  modify_table_body(
    ~ .x %>% 
      dplyr::left_join(
        weighted_mean, 
        by = c("row_type", "label")
      )
  ) |> 
  modify_column_unhide(columns = c("stat_3")) |> 
  modify_column_hide(columns = c("stat_0")) |> 
  modify_header(
    label = "**Characteristic**", 
    stat_3 = "**CVH Score**, Mean (SD)"
  )