R语言crossCarry包CrossGEEKron函数报错：replacement has length zero

问题背景 #

我正在开展一项包含2个周期、2种处理、3个时间点的重复测量交叉研究分析，运行CrossGEEKron函数时触发如下错误，且确认数据格式与演示数据Arterial一致，无法定位原因。

报错信息 #

Error in r_media[ii, ] <- media_temp/(n_suj) : replacement has length zero

运行过程输出：

Beginning Cgee S-function, @(#) geeformula.q 4.13 98/01/27
running glm to get initial regression estimate
(Intercept)     period2  treatment2     Carry_2        time 
   78.71847    -2.05556    -6.94444     1.88889     0.06081 
Error in r_media[ii, ] <- media_temp/(n_suj) : 
  replacement has length zero

所用代码 #

carrydata<- createCarry(data=rop, treatment= "treatment", period= "period", id="sno")
data<- carrydata$data
carry<- carrydata$carryover

model<- CrossGEEKron(response= "syst", treatment = "treatment", 
         period = "period", id= "sno", time = "time",
         carry= "Carry_2", data=data, correlation="ar1")

数据集示例 #

time   babysuhidno        sno         syst diast    hr   period    treatment
                                
1   0   104944203   104944203   82  36  172 1   2
2   0   104944203   104944203   76  41  171 2   1
3   0   10583361    10583361    84  47  170 1   2
4   0   10583361    10583361    87  42  156 2   1
5   0   105855122   105855122   83  44  153 1   1
6   0   105855122   105855122   65  33  167 2   2
7   0   105855143   105855143   71  32  168 1   1
8   0   105855143   105855143   76  43  165 2   2
9   0   105904967   105904967   53  26  177 1   2
10  0   105904967   105904967   73  52  163 2   1
11  20  104944203   104944203   81  40  174 1   2
12  20  104944203   104944203   81  51  170 2   1
13  20  10583361    10583361    85  39  181 1   2
14  20  10583361    10583361    83  57  157 2   1
15  20  105855122   105855122   80  36  165 1   1
16  20  105855122   105855122   61  31  171 2   2
17  20  105855143   105855143   72  33  165 1   1
18  20  105855143   105855143   76  41  162 2   2
19  20  105904967   105904967   54  27  172 1   2
20  20  105904967   105904967   82  42  152 2   1
21  35  104944203   104944203   75  33  177 1   2
22  35  104944203   104944203   81  42  159 2   1
23  35  10583361    10583361    80  37  173 1   2
24  35  10583361    10583361    83  48  153 2   1
25  35  105855122   105855122   88  45  156 1   1
26  35  105855122   105855122   72  43  174 2   2
27  35  105855143   105855143   85  66  154 1   1
28  35  105855143   105855143   75  42  133 2   2
29  35  105904967   105904967   62  32  177 1   2
30  35  105904967   105904967   71  41  141 2   1

解决方案 #

这个错误通常是CrossGEEKron函数无法正确识别受试者的重复测量结构导致的，可按以下步骤排查修复：

• 转换time变量类型：确保time为因子类型，函数需要时间点作为分类变量构建相关结构，执行代码：

  data$time <- as.factor(data$time)
  

• 检查时间点完整性：确认每个受试者的所有周期、处理组合下，3个时间点记录完整，用代码排查缺失：

  library(dplyr)
  data %>%
    group_by(sno, period, treatment) %>%
    summarise(n_time = n_distinct(time)) %>%
    filter(n_time != 3)
  

  若输出非空，补全或剔除不完整的受试者数据。

• 调整相关结构参数：尝试将correlation从"ar1"改为"exchangeable"，ar1结构对测量顺序要求更严格，换用exchangeable可绕过部分场景的错误：

  model<- CrossGEEKron(response= "syst", treatment = "treatment", 
           period = "period", id= "sno", time = "time",
           carry= "Carry_2", data=data, correlation="exchangeable")
  

• 验证carry变量有效性：检查Carry_2无缺失值或异常水平：

  table(data$Carry_2, useNA = "always")
  

  若存在异常，重新运行createCarry并核对参数。

内容的提问来源于stack exchange，提问作者Vivek Kumar