为每个因子水平级别添加频率计数列，并调整数据框的形状

Question

I have a dataframe like so:

 df<- data.frame(region = c("1","1","1","1","1","2","2"),
            loc = c("104","104","104","105","105","106","107"), 
            plntsp = c("A","A", "B", "C", "C", "E", "F"), 
            lepsp = c("Z","Z", "Y", "W", "X", "T", "T"))

And I want to: 1) find the frequencies of plntsp and lepsp for each region and loc subset. 2)make it a long dataframe where the plantsp and lepsp columns are collapsed into one column titled sp . And the new count columns are collapsed into one count column called freq .

output<- data.frame(region = c("1","1","1","1","2","1","1","1","2","2","2"),
loc = c("104","104","105","106","107","104","104","105","105","106","107"), 
sp = c("A","B", "C", "E", "F", "Z", "Y","W", "X", "T", "T"), 
freq = c("2","1", "2", "1", "1", "2", "1", "1", "1", "1", "1"))

I have tried:

df<- 
group_by(region,loc) %>%
summarise(freq1= length(unique(plantsp), freq2= length(unique(lepsp))
mutate(sp= df$plantsp &df$lepsp, freq= df$freq1 &df$freq2)

Answer 1

aggregate would be one option

rbind(aggregate(list(freq = seq_along(df$plntsp)),
                by = list(region = df$region,loc = df$loc, sp = df$plntsp),
                FUN = length),
      aggregate(list(freq = seq_along(df$plntsp)),
                by = list(region = df$region, loc = df$loc, sp = df$lepsp),
                FUN = length))
#   region loc sp freq
#1       1 104  A    2
#2       1 104  B    1
#3       1 105  C    2
#4       2 106  E    1
#5       2 107  F    1
#6       2 106  T    1
#7       2 107  T    1
#8       1 105  W    1
#9       1 105  X    1
#10      1 104  Y    1
#11      1 104  Z    2

Or use melt of reshape2 before using aggregate

library(reshape2)
opt = melt(data = df, id.vars = c("region", "loc"))
#Warning message:
#attributes are not identical across measure variables; they will be dropped 
aggregate(list(freq=opt$value), opt[c("region","loc","value")], FUN = length)
#   region loc value freq
#1       1 104     A    2
#2       1 104     B    1
#3       1 105     C    2
#4       2 106     E    1
#5       2 107     F    1
#6       2 106     T    1
#7       2 107     T    1
#8       1 105     W    1
#9       1 105     X    1
#10      1 104     Y    1
#11      1 104     Z    2

Answer 2

Using tidyverse :

library(tidyverse)
df %>% 
  gather(key, sp, plntsp, lepsp) %>%
  group_by(region, loc, sp) %>%
  count(.) %>%
  rename(x=n)

   region    loc    sp     x
 1      1    104     A     2
 2      1    104     B     1
 3      1    104     Y     1
 4      1    104     Z     2
 5      1    105     C     2
 6      1    105     W     1
 7      1    105     X     1
 8      2    106     E     1
 9      2    106     T     1
10      2    107     F     1
11      2    107     T     1

Answer 3

This data.table solution follows the advice from thelatemail to reshape to long format first and then to count appearances.

The melt() function to reshape data from wide to long format is available from two packages: reshape2 and data.table . I prefer the latter for performance reasons and the concise syntax:

library(data.table)
id_vars = c("region", "loc")
melt(setDT(df), id.vars = id_vars, value.name = "sp")[, .(freq = .N), c(id_vars, "sp")]

  region loc sp freq 1: 1 104 A 2 2: 1 104 B 1 3: 1 105 C 2 4: 2 106 E 1 5: 2 107 F 1 6: 1 104 Z 2 7: 1 104 Y 1 8: 1 105 W 1 9: 1 105 X 1 10: 2 106 T 1 11: 2 107 T 1 

Note that the columns have been renamed as requested by the OP. For comparison with the other answers posted the far, the code is even more condensed without renaming the columns:

melt(setDT(df), id.vars = id_vars)[, .N, c(id_vars, "value")]