使用circumplex包分析环形数据时purrr::map_dbl报错的解决咨询

报错信息 #

Error in **purrr::map_dbl()**:
ℹ In index: 1.
ℹ With name: t1.
Caused by error in **quantile.default()**:
! missing values and NaN's not allowed if 'na.rm' is FALSE

执行代码 #

res.a1 <- ssm_analyze(.data = df.area1, 
    scales = c(a.pos1, v.neg1, a.neg1, v.pos1),                           
    angles = c(90, 180, 270, 360), measures = c(safe1, inter1), 
    measures_labels = c('Safety', 'Interaction')) 
summary(res.a1)

回溯信息 #

Backtrace:
▆

 1. ├─circumplex::ssm_analyze(...)
 2. │ └─circumplex:::ssm_analyze_corrs(...)
 3. │   └─circumplex:::ssm_bootstrap(...)
 4. │     └─... %>% dplyr::select(-fit_lci)
 5. ├─dplyr::select(., -fit_lci)
 6. ├─circumplex (local) reshape_params(., suffix = "lci")
 7. │ └─... %>% tibble::as_tibble(nrow = nrow(.))
 8. ├─tibble::as_tibble(., nrow = nrow(.))
 9. ├─base::`colnames<-`(...)
10. │ └─base::is.data.frame(x)
11. ├─base::matrix(., ncol = 6, byrow = TRUE)
12. └─purrr::map_dbl(., .f = quantile, probs = ((1 - interval)/2))
13. └─purrr:::map_("double", .x, .f, ..., .progress = .progress)
14.     ├─purrr:::with_indexed_errors(...)
15.     │ └─base::withCallingHandlers(...)
16.     ├─purrr:::call_with_cleanup(...)
17.     ├─stats (local) .f(.x[[i]], ...)
18.     └─stats:::quantile.default(.x[[i]], ...)
19.       └─base::stop("missing values and NaN's not allowed if 'na.rm' is FALSE")

Run rlang::last_trace(drop = FALSE) to see 4 hidden frames.

数据情况 #

已确认所有变量观测数均为94，数据样本如下：

id  v.neg1           v.pos1           a.neg1    a.pos1       inter1   safe1

1   -0.1031421  -1.7056656  -1.6342515  -0.2935107  6   5

2   -0.1031421  -0.7625329  0.9694712   -0.2935107  5   7

3   -0.1031421  0.1805999   0.3185405   -0.2935107  7   8

4   9.5922176   1.1237326   0.3185405   -0.2935107  3   8

5   -0.1031421  0.1805999   0.3185405   -0.2935107  5   8

6   -0.1031421  1.1237326   0.9694712   -0.2935107  8   8

7   -0.1031421  -0.7625329  -1.6342515  -0.2935107  5   5

8   -0.1031421  -1.7056656  -1.6342515  -0.2935107  4   5

9   -0.1031421  0.1805999   -1.6342515  -0.2935107  6   7

10  -0.1031421  2.0668654   0.9694712   -0.2935107  9   9

解决方案 #

从回溯信息看，报错出现在bootstrap抽样后的分位数计算环节。原数据无显性缺失，但抽样过程或内部计算可能产生NaN/缺失值，可尝试以下方法解决：

• 清理隐形缺失值：先排查并替换数据中的Inf或NaN，再删除含缺失值的行：

  df.area1 <- df.area1 %>% 
    mutate(across(everything(), ~ifelse(is.infinite(.) | is.nan(.), NA, .))) %>% 
    drop_na()
  

• 调整bootstrap参数：关闭bootstrap或减少抽样次数，避免异常样本引发错误：

  # 关闭bootstrap
  res.a1 <- ssm_analyze(.data = df.area1, 
      scales = c(a.pos1, v.neg1, a.neg1, v.pos1),                           
      angles = c(90, 180, 270, 360), measures = c(safe1, inter1), 
      measures_labels = c('Safety', 'Interaction'),
      bootstrap = FALSE)
  
  # 或减少抽样次数
  res.a1 <- ssm_analyze(.data = df.area1, 
      scales = c(a.pos1, v.neg1, a.neg1, v.pos1),                           
      angles = c(90, 180, 270, 360), measures = c(safe1, inter1), 
      measures_labels = c('Safety', 'Interaction'),
      n_boot = 100)
  

• 修正环形角度设置：环形数据通常用0代替360，避免计算冲突：

  angles = c(90, 180, 270, 0)
  

• 检查变量方差：样本中a.pos1全部为同一值，方差为0会导致相关系数计算产生NaN，进而触发报错。运行以下代码确认：

  sapply(df.area1[, c("a.pos1", "v.neg1", "v.pos1", "a.neg1", "inter1", "safe1")], var)
  

  若存在方差为0的变量，需重新检查数据预处理流程，或根据研究设计调整变量选择。

内容的提问来源于stack exchange，提问作者Ola