Question

我有一个缺少观察值的数据集。我的目标是用之前的13个值的平均值来估算每个变量中的缺失值。如果在第13个观测值之前存在缺失值，则应使用之前的平均值来估算该变量。我不确定该怎么做。

请使用以下内容复制我的数据集。非常感谢您的帮助。

df1 <- structure(list(V1 = c(276.12, 53.4, 20.64, 181.8, 216.96, 10.44, 
69, 144.24, 10.32, 239.76, 79.32, 257.64, 28.56, 117, 244.92, 
234.48, NA, 337.68, 83.04, 176.76, 262.08, 284.88, 15.84, NA, 
74.76, 315.48, 171.48, 288.12, 298.56, 84.72, 351.48, 135.48, 
NA, 318.72, 114.84, 348.84, 320.28, 89.64, 51.72, 273.6, 243, 
212.4, 352.32, 248.28, NA, 210.12, 107.64, 287.88, 272.64, 80.28, 
239.76, 120.48, 259.68, 219.12, 315.24, 238.68, 8.76, 163.44, 
252.96), V2 = c(45.36, 47.16, 55.08, 49.56, 12.96, 58.68, 39.36, 
NA, 2.52, 3.12, 6.96, 28.8, NA, 9.12, 39.48, 57.24, 43.92, 47.52, 
24.6, 28.68, 33.24, 6.12, 19.08, 20.28, 15.12, 4.2, 35.16, NA, 
32.52, 19.2, 33.96, 20.88, 1.8, 24, 1.68, NA, 52.56, 59.28, 32.04, 
45.24, 26.76, 40.08, 33.24, 10.08, 30.84, 27, 11.88, 49.8, 18.96, 
14.04, 3.72, 11.52, 50.04, 55.44, 34.56, NA, 33.72, 23.04, 59.52
)), class = "data.frame", row.names = c(NA, -59L))

Answer 1

您可以使用zoo::rollapply来计算13个值的平均值：

mean13 = zoo::rollapply(
    df1$V1, 
    13, 
    function(x) { 
    mean(na.omit(x)) 
    }, 
    align = "right", 
    fill = NA, 
    partial = TRUE
)
df1$V1_prev_mean = c(df1$V1[1], head(mean13, -1))
df1$V1 = ifelse(is.na(df1$V1), df1$V1_prev_mean, df1$V1)

输出：

         V1    V2 V1_prev_mean
1  276.1200 45.36     276.1200
2   53.4000 47.16     276.1200
3   20.6400 55.08     164.7600
4  181.8000 49.56     116.7200
5  216.9600 12.96     132.9900
6   10.4400 58.68     149.7840
7   69.0000 39.36     126.5600
8  144.2400    NA     118.3371
9   10.3200  2.52     121.5750
10 239.7600  3.12     109.2133
11  79.3200  6.96     122.2680
12 257.6400 28.80     118.3636
13  28.5600    NA     129.9700
14 117.0000  9.12     122.1692
15 244.9200 39.48     109.9292
16 234.4800 57.24     124.6615
17 141.1108 43.92     141.1108  # <- this row filled
18 337.6800 47.52     137.7200
19  83.0400 24.60     147.7800
20 176.7600 28.68     153.8300

推算缺失值，并将其与之前的13个值进行平均

1 个答案: