R:使用带有权重的 reshape2::acast 聚合

Question

我的目标是使用 reshape2::acast() 创建一个二维数组，其中聚合函数是加权平均值。这是一个便宜的加权平均函数:

wmean <- function(x,w=rep(1,length(x))){
    if (length(x)==0){
        return(NA)
    }
    sum(x * w, na.rm = TRUE) / sum(w, na.rm = TRUE)
}

这是一些数据(100 行，4 列)。

Dat <- structure(list(cafloor2 = c(62, 62, 64, 60, 62, 64, 70, 72, 74, 
76, 78, 60, 62, 64, 66, 68, 60, 60, 62, 64, 66, 60, 62, 62, 62, 
64, 66, 68, 70, 60, 62, 64, 60, 62, 64, 66, 68, 70, 62, 62, 64, 
68, 60, 62, 64, 66, 60, 62, 60, 62, 64, 60, 62, 60, 62, 64, 62, 
64, 66, 68, 70, 72, 74, 76, 62, 60, 62, 64, 60, 62, 64, 66, 68, 
60, 62, 64, 66, 68, 72, 74, 78, 80, 60, 60, 62, 64, 66, 60, 62, 
60, 64, 60, 62, 64, 66, 68, 82, 84, 88, 60), tafloor2 = c(4, 
2, 0, 4, 2, 0, 8, 6, 4, 2, 0, 8, 6, 4, 2, 0, 0, 6, 4, 4, 0, 2, 
0, 0, 12, 10, 10, 6, 4, 6, 4, 0, 12, 10, 6, 4, 4, 0, 8, 6, 4, 
2, 12, 10, 6, 6, 2, 0, 6, 4, 2, 4, 2, 4, 2, 0, 14, 12, 10, 8, 
6, 4, 2, 0, 0, 4, 2, 0, 8, 6, 2, 2, 0, 8, 6, 4, 2, 0, 8, 6, 2, 
0, 8, 6, 4, 4, 2, 2, 0, 2, 0, 8, 6, 4, 2, 0, 8, 6, 2, 2), srh = c(4L, 
5L, 3L, 1L, 2L, 3L, 4L, 4L, 4L, 4L, 3L, 4L, 5L, 5L, 5L, 5L, 3L, 
5L, 5L, 5L, 5L, 5L, 4L, 5L, 2L, 2L, 3L, 2L, 2L, 4L, 4L, 4L, 1L, 
1L, 1L, 1L, 2L, 2L, 4L, 5L, 5L, 3L, 5L, 4L, 4L, 4L, 3L, 2L, 1L, 
2L, 2L, 4L, 3L, 4L, 4L, 4L, 3L, 2L, 2L, 3L, 2L, 3L, 3L, 3L, 2L, 
5L, 5L, 5L, 2L, 4L, 3L, 2L, 1L, 4L, 3L, 3L, 4L, 4L, 3L, 3L, 4L, 
2L, 4L, 5L, 5L, 4L, 5L, 3L, 4L, 4L, 3L, 4L, 3L, 4L, 4L, 4L, 3L, 
4L, 4L, 4L), p_wt2 = c(3065, 3121, 3390, 6122, 5747, 6488, 2045, 
1880, 3083, 2326, 2288, 1563, 1789, 1676, 1536, 1586, 1517, 2231, 
2420, 2066, 2108, 2015, 2031, 6239, 4718, 4460, 4735, 5183, 5359, 
5084, 5400, 5280, 4439, 4714, 4939, 5359, 5143, 5298, 2298, 2496, 
2650, 2719, 2025, 2341, 1990, 2015, 5114, 5554, 4304, 3824, 3962, 
4718, 5253, 1475, 1556, 1657, 1475, 1556, 1657, 1765, 1736, 1845, 
1760, 1751, 1291, 1749, 1845, 1770, 1556, 1657, 1765, 1736, 1747, 
1324, 1453, 1711, 1898, 1830, 1563, 1518, 971, 996, 1326, 1194, 
1293, 1447, 1347, 1536, 1628, 1749, 1845, 1260, 1456, 1165, 1122, 
1156, 1342, 2063, 1643, 1491)), .Names = c("cafloor2", "tafloor2", 
"srh", "p_wt2"), row.names = c(NA, 100L), class = "data.frame")

现在，我可以使用以下方法将矩阵作为未加权均值:

acast(Dat,cafloor2~tafloor2, value.var = "srh", 
        fun.aggregate = mean, na.rm=TRUE)

但我想要加权平均值:

acast(Dat,cafloor2~tafloor2, value.var = "srh", 
            fun.aggregate = wmean, w = "p_wt2")

根据需要从 Dat 中获取权重。这给出了错误:

Error in x * w : non-numeric argument to binary operator

也许我今天脑子有点迟钝，但我一直在思考如何进行。当然欢迎使用非acast() 解决方案。谢谢!

Answer 1

我似乎不太明白如何使用 acast 获得正确的组权重向量。鉴于此，我会通过简单地先按组计算汇总统计数据然后转换结果来解决这个问题。下面是一个使用 dplyr 计算加权平均值的示例。

require(dplyr)
Dat %>% group_by(cafloor2, tafloor2) %>% 
    summarise(wsrh = wmean(srh, p_wt2)) %>%
    acast(cafloor2 ~ tafloor2, value.var = "wsrh")

您可以使用 tidyr 中的 spread 而不是 acast，尽管输出格式略有不同，您可能更喜欢 acast 输出。

require(tidyr)
Dat %>% group_by(cafloor2, tafloor2) %>% 
    summarise(wsrh = wmean(srh, p_wt2)) %>% 
    spread(tafloor2, wsrh)