Question

我有一个区，县和年的数据集。如果某个地区/县组合在任何一年发生，我希望每年都能实现这种组合。以下是我想到的两种方法。第一种方法使用函数来创建区，县和年的组合，并且只需要六行代码。底层方法使用paste，expand.grid和strsplit的组合，并且更加复杂/复杂。

可能比上述方法更有效。例如，有没有办法使用可能只有1或2行代码来实现区/县/年组合的expand.grid？

感谢您的任何建议。我的职能可以胜任，但这个问题对我来说是一个学习的机会。我更喜欢基地R。

以下是示例数据集：

df.1 <- read.table(text = '
    state    district    county   year   apples
       AA          EC        A    1980     100
       AA          EC        B    1980      10
       AA          EC        C    1980     150
       AA           C        G    1980     200
       AA           C    other    1980      20
       AA           C        I    1980     250
       AA          WC        R    1980     300
       AA          WC        S    1980      30
       AA          WC     other   1980     350
       AA          EC        A    1999    1100
       AA          EC        D    1999     110
       AA          EC        E    1999    1150
       AA           C        H    1999    1200
       AA           C        I    1999     120
       AA           C        J    1999    1250
       AA          WC        R    1999    1300
       AA          WC    other    1999     130
       AA          WC        T    1999    1350
', header=TRUE, stringsAsFactors = FALSE)

这是期望的结果：

desired.result <- read.table(text = '
   state district county  year apples
      AA        C      G  1980    200
      AA        C      H  1980     NA
      AA        C      I  1980    250
      AA        C      J  1980     NA
      AA        C  other  1980     20
      AA       EC      A  1980    100
      AA       EC      B  1980     10
      AA       EC      C  1980    150
      AA       EC      D  1980     NA
      AA       EC      E  1980     NA
      AA       WC  other  1980    350
      AA       WC      R  1980    300
      AA       WC      S  1980     30
      AA       WC      T  1980     NA
      AA        C      G  1999     NA
      AA        C      H  1999   1200
      AA        C      I  1999    120
      AA        C      J  1999   1250
      AA        C  other  1999     NA
      AA       EC      A  1999   1100
      AA       EC      B  1999     NA
      AA       EC      C  1999     NA
      AA       EC      D  1999    110
      AA       EC      E  1999   1150
      AA       WC  other  1999    130
      AA       WC      R  1999   1300
      AA       WC      S  1999     NA
      AA       WC      T  1999   1350
', header=TRUE, stringsAsFactors = FALSE)

这是迄今为止我最直接的解决方案，它使用一个函数来表示每年的每个区/县组合：

my.unique.function <- function(year) {
     my.unique     <- data.frame(unique(df.1[, c('state', 'district', 'county')]), year)
     return(my.unique = my.unique)
}

years <- as.data.frame(unique(df.1[, 'year']))
my.unique.output <- apply(years, 1, function(x) {my.unique.function(x)})
my.unique.output2 <- do.call(rbind.data.frame, my.unique.output)

desired.result2 <- merge(df.1, my.unique.output2, by = c('state', 'year', 'district', 'county'), all=TRUE)

# compare output with a priori desired result
desired.result <- desired.result[order(desired.result$state, desired.result$year, desired.result$district, desired.result$county),]
all.equal(desired.result[,c(1,4,2,3,5)], desired.result2[,1:5])

这是我最初的，更复杂的解决方案：

# find unique combinations of district and county
my.unique     <- unique(df.1[, c('district', 'county')])

# paste district and county together
my.unique$x   <- apply( my.unique[ , c('district', 'county') ] , 1 , paste , collapse = "-" )

# represent each district/county combination for each year
expand.unique      <- expand.grid(my.unique$x, year = c(1980, 1999))
expand.unique$Var1 <- as.character(expand.unique$Var1)

# split combined district/county combinations into two columns
expand.unique$Var1b <- sub('-', ' ', expand.unique$Var1)
unique.split        <- strsplit(expand.unique$Var1b, ' ')
unique.splits       <- matrix(unlist(unique.split), ncol=2, byrow=TRUE, dimnames = list(NULL, c("district", "county")))

# create template prior to merging with original data set
state <- 'AA'
desired.resultb <- data.frame(state, expand.unique, unique.splits)

# merge template with original data set so missing observations are present if a county is not included for a given year
desired.resultc <- merge(df.1, desired.resultb, by = c('state', 'year', 'district', 'county'), all=TRUE)
desired.resultc

# compare output with a priori desired result
desired.result <- desired.result[order(desired.result$state, desired.result$year, desired.result$district, desired.result$county),]
all.equal(desired.result[,c(1,4,2,3,5)], desired.resultc[,1:5])

Answer 1

#find all (unique) state-district-county combinations
df.2 <- unique(df.1[,c("state","district","county")])

#find all (unique) years
df.3 <- unique(df.1[,"year",drop=FALSE])

#Cartesian product of combinations
df.4 <- merge(df.2,df.3)

#merge this with the original data frame,
#leaving NA's for unmatched parts in df.4
merge(df.1,df.4,all=TRUE)

Answer 2

以下是使用expand.grid的解决方案，只需要四行。但是，我更喜欢Blue Magister的回答。

my.template  <- expand.grid( unique(paste(df.1$state, df.1$district, df.1$county, sep= ' ')), year = unique(df.1$year))

my.template2 <- data.frame(do.call(rbind, strsplit(as.character(my.template$Var1), ' ')), year = my.template$year)

colnames(my.template2) <- names(df.1)[1:4]

desired.result2 <- merge(df.1, my.template2, all=TRUE)

当一个变量实际上是两列时，expand.grid

2 个答案: