在R中重塑从长到宽的数据集时有条件地填充缺少的值

我正在根据不同质量的多个数据集，为一系列年份和国家构建完整的指标时间表

使用

reformae2

我将这些数据集“融合”到一个数据帧中

示例数据集：

d <- structure(list(cntry = structure(c(1L, 1L, 1L, 2L, 2L, 3L, 3L, 
1L, 1L, 2L, 2L, 3L, 3L, 1L, 1L, 2L, 2L, 3L, 3L), .Label = c("BE", 
"DE", "GE"), class = "factor"), year = c(1960L, 1970L, 1980L, 
1960L, 1970L, 1960L, 1970L, 1960L, 1970L, 1960L, 1970L, 1960L, 
1970L, 1960L, 1970L, 1960L, 1970L, 1970L, 1980L), indicator = c(5.5, 
1.2, 1.5, NA, 1.4, NA, NA, 5.5, 1.2, 2.3, 1.4, NA, 1.4, NA, NA, 
2.3, 1.4, 1.4, NA), sex = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = "male", class = "factor"), 
    source = structure(c(2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 
    3L, 3L, 3L, 3L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("Council", 
    "Eurostat", "OECD"), class = "factor")), .Names = c("cntry", 
"year", "indicator", "sex", "source"), class = "data.frame", row.names = c(NA, 
-19L))


d
#    cntry year indicator  sex   source
# 1     BE 1960       5.5 male Eurostat
# 2     BE 1970       1.2 male Eurostat
# 3     BE 1980       1.5 male Eurostat
# 4     DE 1960        NA male Eurostat
# 5     DE 1970       1.4 male Eurostat
# 6     GE 1960        NA male Eurostat
# 7     GE 1970        NA male Eurostat
# 8     BE 1960       5.5 male     OECD
# 9     BE 1970       1.2 male     OECD
# 10    DE 1960       2.3 male     OECD
# 11    DE 1970       1.4 male     OECD
# 12    GE 1960        NA male     OECD
# 13    GE 1970       1.4 male     OECD
# 14    BE 1960        NA male  Council
# 15    BE 1970        NA male  Council
# 16    DE 1960       2.3 male  Council
# 17    DE 1970       1.4 male  Council
# 18    GE 1970       1.4 male  Council
# 19    GE 1980        NA male  Council

并可选择（或直接）转换为宽格式：

# cntry  sex 1960 1970 1980
#    BE male  5.5  1.2  1.5
#    DE male  2.3  1.4  NA
#    GE male   NA  1.4  NA

---

我不确定这是否符合您的所有期望，但听起来您正在寻找以下内容：

toMerge <- expand.grid(cntry = c("BE", "DE", "GE"), 
                       year = c(1960, 1970, 1980), 
                       source = c("Eurostat", "OECD", "Council"), 
                       sex = "male")
d2 <- merge(d, toMerge, all = TRUE)

d2$source <- factor(d2$source, c("Council", "OECD", "Eurostat"), ordered=TRUE)
d2 <- d2[order(d2$source, decreasing=TRUE), ]
Rank <- with(d2, ave(indicator, d2[c("cntry", "year", "sex")], 
                 FUN = function(x) rank(x, ties.method="first", na.last=TRUE)))
D <- d2[Rank == 1, ]
D
#    cntry year  sex   source indicator
# 2     BE 1960 male Eurostat       5.5
# 5     BE 1970 male Eurostat       1.2
# 8     BE 1980 male Eurostat       1.5
# 14    DE 1970 male Eurostat       1.4
# 17    DE 1980 male Eurostat        NA
# 20    GE 1960 male Eurostat        NA
# 26    GE 1980 male Eurostat        NA
# 12    DE 1960 male     OECD       2.3
# 24    GE 1970 male     OECD       1.4

library(reshape2)
dcast(D, cntry ~ year, value.var="indicator")
#   cntry 1960 1970 1980
# 1    BE  5.5  1.2  1.5
# 2    DE  2.3  1.4   NA
# 3    GE   NA  1.4   NA

---

toMerge这里是另一个选项：

library(reshape2)
d$source <- factor(d$source, levels=c('Eurostat', 'OECD', 'Council'))
d2 <- d[1:4]
d2[[3]] <- lapply(split(d, 1:nrow(d)), `[`, c(3, 5))
dcast(
  d2, cntry + sex ~ year, value.var="indicator", 
  fun.aggregate=function(x) {
    if(!length(x)) return(NA_real_)
    xs <- do.call(rbind, x)
    xs <- xs[complete.cases(xs), ]
    if(nrow(xs)) xs[order(as.numeric(xs$source)), "indicator"][[1L]] else NA_real_
} )

注：我在“欧盟统计局”值上加了100，以使其与其他值区别开来，因为在这个样本集中，它们似乎是相等的

基本上，我们通过将
指标
列转换为包含指标和来源的列表项列来作弊，然后使用
fun.aggregate
从每个组中选择来源值最低的项目（注意，我们重置了因子，以便最理想的来源具有最低的级别）
 也许以下方法也能奏效：

library(reshape2)
x <- melt(d,id.vars=c("cntry","year","source","sex"))
y <- dcast(x,cntry+year+sex ~ source)
y$selected.value <- ifelse(is.na(y$Eurostat),yes=ifelse(is.na(y$OECD),yes=y$Council,no=y$OECD),no=y$Eurostat)
dcast(y,cntry + sex ~ year)

假设数据是按照您要求的顺序排列的，也就是说，
源
列首先由
欧盟统计局
排序，然后由
经合组织
排序，然后由
理事会
排序，我将继续使用
数据。表
如下：

require(data.table) # >= v1.9.0
setDT(d) # converts data.frame to data.table by reference
dcast.data.table(d, cntry + sex ~ year, value.var="indicator", 
 subset=.(!duplicated(d, by=c("cntry", "year", "indicator")) & !is.na(indicator)))

#    cntry  sex 1960 1970 1980
# 1:    BE male  5.5  1.2  1.5
# 2:    DE male  2.3  1.4   NA
# 3:    GE male   NA  1.4   NA

简洁的解决方案
library(reshape2)
x <- melt(d,id.vars=c("cntry","year","source","sex"))
y <- dcast(x,cntry+year+sex ~ source)
y$selected.value <- ifelse(is.na(y$Eurostat),yes=ifelse(is.na(y$OECD),yes=y$Council,no=y$OECD),no=y$Eurostat)
dcast(y,cntry + sex ~ year)

y$selected.source <- ifelse(is.na(y$Eurostat),yes=ifelse(is.na(y$OECD),yes="Council",no="OECD"),no="Eurostat")

require(data.table) # >= v1.9.0
setDT(d) # converts data.frame to data.table by reference
dcast.data.table(d, cntry + sex ~ year, value.var="indicator", 
 subset=.(!duplicated(d, by=c("cntry", "year", "indicator")) & !is.na(indicator)))

#    cntry  sex 1960 1970 1980
# 1:    BE male  5.5  1.2  1.5
# 2:    DE male  2.3  1.4   NA
# 3:    GE male   NA  1.4   NA