使用gtsummary的by参数时，如何排除无观测值的列？

问题出在career_stage_s1_v2大概率是因子类型，即便筛选后，原因子的8个水平依然被保留，导致gtsummary生成表格时会显示所有水平列。以下两种方法可解决该问题：

方法1：重置因子水平（推荐，从数据源头处理） #

在筛选后重置career_stage_s1_v2的因子水平，仅保留当前数据中存在的取值：

table10 <- select(df, career_stage_s1_v2, mhcsf_emotional_wb, mhcsf_social_wb, mhcsf_psychological_wb) %>%
  filter(career_stage_s1_v2 %in% c("Student", "Currently practicing") & 
           (mhcsf_emotional_wb != "Incomplete" | mhcsf_social_wb != "Incomplete" | mhcsf_psychological_wb != "Incomplete")) %>%
  # 重置因子水平，仅保留有观测的取值
  mutate(career_stage_s1_v2 = factor(career_stage_s1_v2)) %>%
  tbl_summary(
    by = career_stage_s1_v2,
    label = list(career_stage_s1_v2 ~ "Student, Currently practicing"),
    sort = all_categorical() ~ "frequency"
  ) %>%
  modify_caption("**Well Being - emotional, social, or psychological well-being domains by career stage**") %>%
  bold_labels()
table10 

方法2：用gtsummary内置函数过滤列 #

若不想修改原数据，可在生成表格后，通过modify_column_filter过滤掉观测数为0的列：

table10 <- select(df, career_stage_s1_v2, mhcsf_emotional_wb, mhcsf_social_wb, mhcsf_psychological_wb) %>%
  filter(career_stage_s1_v2 %in% c("Student", "Currently practicing") & 
           (mhcsf_emotional_wb != "Incomplete" | mhcsf_social_wb != "Incomplete" | mhcsf_psychological_wb != "Incomplete")) %>%
  tbl_summary(
    by = career_stage_s1_v2,
    label = list(career_stage_s1_v2 ~ "Student, Currently practicing"),
    sort = all_categorical() ~ "frequency"
  ) %>%
  # 过滤掉观测数为0的分组列
  modify_column_filter(columns = all_stat_cols(), ~ !is.na(.x) & .x > 0) %>%
  modify_caption("**Well Being - emotional, social, or psychological well-being domains by career stage**") %>%
  bold_labels()
table10 

内容的提问来源于stack exchange，提问作者Greg Kawchuk