R 从长到宽,具有自动虚拟创建和多个值列
我坐在一个数据框前面,看起来像这样:R 从长到宽,具有自动虚拟创建和多个值列,r,dataframe,dplyr,tidyr,tidyverse,R,Dataframe,Dplyr,Tidyr,Tidyverse,我坐在一个数据框前面,看起来像这样: country year Indicator a b c 48996 US 2003 var1 NA NA NA 16953 FR 1988 var2 NA 10664.920 NA 22973 FR 1943 var3 NA 5774.334 NA
country year Indicator a b c
48996 US 2003 var1 NA NA NA
16953 FR 1988 var2 NA 10664.920 NA
22973 FR 1943 var3 NA 5774.334 NA
8760 CN 1995 var4 8804.565 NA 12750.31
47795 US 2012 var5 NA NA NA
30033 GB 1969 var6 NA 29631.362 NA
25796 FR 1921 var7 NA 14004.520 NA
39534 NL 1941 var8 NA NA NA
42255 NZ 1969 var8 NA NA NA
7249 CN 1995 var9 50635.862 NA 75260.56
我想做的基本上是一个从长到宽的转换,使用指示符
作为关键变量。我通常使用tidyr
包中的spread()
。但是,不幸的是,spread()
不接受多个值列(在本例中,a
、b
和c
),并且它没有完全实现我想要实现的目标:
指示器的条目设置为新列
a
、b
和c
country year var1 [...] var4 [...] var9 dummy.a dummy.b dummy.c
CN 1995 NA 8804.565 50635.862 1 0 0
CN 1995 NA 12750.31 75260.56 0 0 1
由于我的原始数据帧是58.162x119,我希望不包含大量手动工作的内容:-)
我希望我清楚地知道我想要实现什么。谢谢你的帮助
可以使用以下代码再现上述数据帧:
structure(list(country = c("US", "FR", "FR", "CN", "US", "GB",
"FR", "NL", "NZ", "CN"), year = c(2003L, 1988L, 1943L, 1995L,
2012L, 1969L, 1921L, 1941L, 1969L, 1995L), Indicator = structure(c(1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 8L, 9L), .Label = c("var1", "var2",
"var3", "var4", "var5", "var6", "var7", "var8", "var9", "var10",
"var11", "var12", "var13", "var14", "var15", "var16", "var17",
"var18"), class = "factor"), a = c(NA, NA, NA, 8804.56480733,
NA, NA, NA, NA, NA, 50635.8621327), b = c(NA, 10664.9199219,
5774.33398438, NA, NA, 29631.3618614, 14004.5195312, NA, NA,
NA), c = c(NA, NA, NA, 12750.3056855, NA, NA, NA, NA, NA, 75260.555946
)), .Names = c("country", "year", "Indicator", "a", "b", "c"), row.names = c(48996L,
16953L, 22973L, 8760L, 47795L, 30033L, 25796L, 39534L, 42255L,
7249L), class = "data.frame")
以下是我的解决方案:
require(tidyr)
mydf <- structure(list(country = c("US", "FR", "FR", "CN", "US", "GB",
"FR", "NL", "NZ", "CN"), year = c(2003L, 1988L, 1943L, 1995L,
2012L, 1969L, 1921L, 1941L, 1969L, 1995L), Indicator = structure(c(1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 8L, 9L), .Label = c("var1", "var2",
"var3", "var4", "var5", "var6", "var7", "var8", "var9", "var10",
"var11", "var12", "var13", "var14", "var15", "var16", "var17",
"var18"), class = "factor"), a = c(NA, NA, NA, 8804.56480733,
NA, NA, NA, NA, NA, 50635.8621327), b = c(NA, 10664.9199219,
5774.33398438, NA, NA, 29631.3618614, 14004.5195312, NA, NA,
NA), c = c(NA, NA, NA, 12750.3056855, NA, NA, NA, NA, NA, 75260.555946
)), .Names = c("country", "year", "Indicator", "a", "b", "c"), row.names = c(48996L,
16953L, 22973L, 8760L, 47795L, 30033L, 25796L, 39534L, 42255L,
7249L), class = "data.frame")
mydf %>% gather(key=newIndicator,value=values, a,b,c) %>% filter(!is.na(values)) %>% spread(key=Indicator,values) %>% mutate(indicatorValues=1) %>% spread(newIndicator,indicatorValues,fill=0)
以下是我的解决方案:
require(tidyr)
mydf <- structure(list(country = c("US", "FR", "FR", "CN", "US", "GB",
"FR", "NL", "NZ", "CN"), year = c(2003L, 1988L, 1943L, 1995L,
2012L, 1969L, 1921L, 1941L, 1969L, 1995L), Indicator = structure(c(1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 8L, 9L), .Label = c("var1", "var2",
"var3", "var4", "var5", "var6", "var7", "var8", "var9", "var10",
"var11", "var12", "var13", "var14", "var15", "var16", "var17",
"var18"), class = "factor"), a = c(NA, NA, NA, 8804.56480733,
NA, NA, NA, NA, NA, 50635.8621327), b = c(NA, 10664.9199219,
5774.33398438, NA, NA, 29631.3618614, 14004.5195312, NA, NA,
NA), c = c(NA, NA, NA, 12750.3056855, NA, NA, NA, NA, NA, 75260.555946
)), .Names = c("country", "year", "Indicator", "a", "b", "c"), row.names = c(48996L,
16953L, 22973L, 8760L, 47795L, 30033L, 25796L, 39534L, 42255L,
7249L), class = "data.frame")
mydf %>% gather(key=newIndicator,value=values, a,b,c) %>% filter(!is.na(values)) %>% spread(key=Indicator,values) %>% mutate(indicatorValues=1) %>% spread(newIndicator,indicatorValues,fill=0)
dt
将是您的原始数据dt2
是最终输出
dt2 <- dt %>%
gather(Parameter, Value, a:c) %>%
spread(Indicator, Value) %>%
mutate(Data = ifelse(rowSums(is.na(.[, paste0("var", 1:9)])) != 9, 1, 0)) %>%
filter(Data != 0) %>%
spread(Parameter, Data, fill = 0) %>%
rename(dummy.a = a, dummy.b = b, dummy.c = c)
dt2%
聚集(参数,值,a:c)%>%
排列(指示器,值)%>%
突变(数据=ifelse(行和(is.na([粘贴0(“var”,1:9)])!=9,1,0))%>%
过滤器(数据!=0)%>%
排列(参数、数据、填充=0)%>%
重命名(dummy.a=a,dummy.b=b,dummy.c=c)
dt
将是您的原始数据dt2
是最终输出
dt2 <- dt %>%
gather(Parameter, Value, a:c) %>%
spread(Indicator, Value) %>%
mutate(Data = ifelse(rowSums(is.na(.[, paste0("var", 1:9)])) != 9, 1, 0)) %>%
filter(Data != 0) %>%
spread(Parameter, Data, fill = 0) %>%
rename(dummy.a = a, dummy.b = b, dummy.c = c)
dt2%
聚集(参数,值,a:c)%>%
排列(指示器,值)%>%
突变(数据=ifelse(行和(is.na([粘贴0(“var”,1:9)])!=9,1,0))%>%
过滤器(数据!=0)%>%
排列(参数、数据、填充=0)%>%
重命名(dummy.a=a,dummy.b=b,dummy.c=c)
Imo,这是一种非常糟糕的数据格式,但您可以像library(data.table)一样到达那里;melt(setDT(DF,keep.rownames=TRUE),id=c(“rn”,“country”,“year”,“Indicator”)[!is.na(value),dcast(.SD,country+year+variable~Indicator)][,dcast(.SD,…~variable,value.var=“variable”,fun=length)]
我认为基于输入的预期不正确。例如,“1983年”的Var4应为8804.565和12750.306。您使用dput
提供的数据集与您的示例不同。例如,在第4行中,是1983年还是1995年?我的错,修复了它。我确实手动更改了一年,以便更清楚地了解我想要实现的目标,但忘记在示例代码中更改它。很抱歉感谢更新的数据集。您能解释一下为什么在虚拟变量a到c
中,第一行CN
是1,0,0
,第二行是0,0,1
?因为基于您的原始数据帧,a
和c
都有这两行的值。依我看,这是一种非常糟糕的数据格式,但您可以像库(data.table)一样到达那里;melt(setDT(DF,keep.rownames=TRUE),id=c(“rn”,“country”,“year”,“Indicator”)[!is.na(value),dcast(.SD,country+year+variable~Indicator)][,dcast(.SD,…~variable,value.var=“variable”,fun=length)]
我认为基于输入的预期不正确。例如,“1983年”的Var4应为8804.565和12750.306。您使用dput
提供的数据集与您的示例不同。例如,在第4行中,是1983年还是1995年?我的错,修复了它。我确实手动更改了一年,以便更清楚地了解我想要实现的目标,但忘记在示例代码中更改它。很抱歉感谢更新的数据集。您能解释一下为什么在虚拟变量a到c
中,第一行CN
是1,0,0
,第二行是0,0,1
?因为基于原始数据帧,a
和c
都有这两行的值。