我有一个数据框，它总结了传递给它的数据框中缺失和非缺失观察的数量[1]。然后我被要求在我拥有的数据中测试两个治疗组之间的差异（我个人不同意这样做的必要性或效用，但这是我被要求做的）。所以我写了一个小函数来做到这一点......
quick.test <- function(x, y){
chisq <- chisq.test(x = x, y = y)
fisher <- fisher.test(x = x, y = y)
results <- cbind(chisq = chisq$statistic,
df = chisq$parameter,
p = chisq$p.value,
fisher = fisher$p.value)
results
}
然后我使用 apply() 将相关列传递给该函数，如下所示...
apply(miss.t1, 1, function(x) quick.test(x[2:3], x[4:5]))
这对于上面指定的 miss.t1 数据框很好，但是我正在使用时间序列数据并且我希望总结三个时间点，所以有 miss.t2 和 miss.t3（每个都在总结数字每个时间点的当前/缺失数据，并已使用 [1] 中描述的函数以相同的方式创建）。
miss.t2 失败并出现以下错误...
apply(miss.t2, 1, function(x) quick.test(x[2:3], x[4:5]))
Error in chisq.test(x = x, y = y) :
'x' and 'y' must have at least 2 levels
我最初的想法是其中一列由于某种原因缺少值，但情况似乎并非如此......
> describe(miss.t2)
miss.t2
5 Variables 171 Observations
--------------------------------------------------------------------------------
variable
n missing unique
171 0 171
lowest : Abtotal Abyn agg_ment agg_phys All.score
highest: z_pf z_re z_rp z_sf z_vt
--------------------------------------------------------------------------------
nmiss.1
n missing unique Mean
171 0 4 8.649
0 (6, 4%), 8 (9, 5%), 9 (153, 89%), 10 (3, 2%)
--------------------------------------------------------------------------------
npresent.1
n missing unique Mean
171 0 4 9.351
8 (3, 2%), 9 (153, 89%), 10 (9, 5%), 18 (6, 4%)
--------------------------------------------------------------------------------
nmiss.2
n missing unique Mean
171 0 4 10.65
0 (6, 4%), 11 (160, 94%), 12 (4, 2%), 13 (1, 1%)
--------------------------------------------------------------------------------
npresent.2
n missing unique Mean
171 0 4 14.35
12 (1, 1%), 13 (4, 2%), 14 (160, 94%), 25 (6, 4%)
--------------------------------------------------------------------------------
我尝试的下一件事是通过获取 head(miss.t2, n=XX) 来尝试 miss.t2 的子集，并且它在第 54 行之前工作正常......
> apply(head(miss.t2, n=53), 1, function(x) quick.test(x[2:3], x[4:5]))
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
[1,] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
[2,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[3,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[4,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
[1,] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
[2,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[3,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[4,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
There were 50 or more warnings (use warnings() to see the first 50)
> apply(head(miss.t2, n=54), 1, function(x) quick.test(x[2:3], x[4:5]))
Error in chisq.test(x = x, y = y) :
'x' and 'y' must have at least 2 levels
> miss.t2[54,]
variable nmiss.1 npresent.1 nmiss.2 npresent.2
54 psq 10 8 11 14
> traceback()
5: stop("'x' and 'y' must have at least 2 levels") at #2
4: chisq.test(x = x, y = y) at #2
3: quick.test(x[2:3], x[4:5])
2: FUN(newX[, i], ...)
1: apply(head(miss.t2, n = 54), 1, function(x) quick.test(x[2:3],
x[4:5]))
与数据框的“底部”类似，最后 26 行被解析得很好，但不是倒数第 27 行......
> apply(tail(miss.t2, n=26), 1, function(x) quick.test(x[2:3], x[4:5]))
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
[1,] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
[2,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[3,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[4,] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
164 165 166 167 168 169 170 171
[1,] 0 0 0 0 0 0 0 0
[2,] 1 1 1 1 1 1 1 1
[3,] 1 1 1 1 1 1 1 1
[4,] 1 1 1 1 1 1 1 1
There were 26 warnings (use warnings() to see them)
> apply(tail(miss.t2, n=27), 1, function(x) quick.test(x[2:3], x[4:5]))
Error in chisq.test(x = x, y = y) :
'x' and 'y' must have at least 2 levels
In addition: Warning message:
In chisq.test(x = x, y = y) : Chi-squared approximation may be incorrect
> miss.t2[118,]
variable nmiss.1 npresent.1 nmiss.2 npresent.2
118 sf16 9 9 11 14
我看不出这两行有什么问题，这意味着它们应该失败，并且上面显示的 traceback() 没有显示任何有用的信息（在我看来）。
任何人都可以就为什么或哪里出了问题提供任何建议吗？
提前谢谢了，
尼尔
编辑：格式化回复文森特 Zoonekynd ...
我选择了 chisq.test(x = x, y = y) 中描述的版本？ （性格）论据。
放置打印语句并显示 x 和 y 的长度会导致相同的错误，但将值和长度显示为...
> miss.t2.res <- data.frame(t(apply(miss.t2, 1, function(x) quick.test(x[2:3], x[4:5]))))
[1] "Your x is : 9" "Your x is : 9"
[1] 2 ### < Length of x
[1] "Your y is : 11" "Your y is : 14"
[1] 2 ### < Length of y
Error in chisq.test(x = x, y = y) : 'x' and 'y' must have at least 2 levels
编辑 2：感谢 Vincent Zoonekynd 的指针，问题是因为两个单元格的计数相同，所以对 chisq.test() 的调用将它们视为因素并折叠它们。解决方案是修改 quick.test() 函数并强制传递给矩阵的参数，所以现在起作用的函数是......
quick.test <- function(x, y){
chisq <- chisq.test(rbind(as.numeric(x), as.numeric(y)))
fisher <- fisher.test(rbind(as.numeric(x), as.numeric(y)))
results <- cbind(chisq = chisq$statistic,
df = chisq$parameter,
p = chisq$p.value,
fisher = fisher$p.value)
results
}
非常感谢文森特的帮助和指点，非常感谢。
[1] http://gettinggeneticsdone.blogspot.co.uk/2011/02/summarize-missing-data-for-all.html