如何基于r中的数字变量的间隔将数据帧分为子组

Question

I have a data frame ( df )that looks like:

mi       chr    gen.pos
m4774   Ch01    0
m4775   Ch01    1.701
m4663   Ch01    5.519
m4777   Ch01    6.5
m4779   Ch01    11.067
m4780   Ch01    11.234
m3933   Ch01    11.449
m4782   Ch01    13.986
m5534   Ch01    119.277
m5536   Ch02    0.036
m5550   Ch02    4.26

the chr column as group, at first, get the intervals of 20 bins of column gen.pos for each group by this code:

len <- as.data.frame(cbind(chr = unique(df$chr), 
  do.call(rbind, tapply(df$gen.pos, df$chr, function(x) {c(min = min(x), max = max(x))}))))
len$interval <- format(round((as.numeric(as.character(len$max))-as.numeric(as.character(len$min)))/20,3),nsmall=3)

so the len data frame is:

chr     min     max     interval
Ch01    0       119.277 5.964
Ch02    0.036   134.249 6.711
Ch03    0.07    93.596  4.676
Ch04    0.392   134.342 6.698
Ch05    0.581   96.842  4.813
Ch06    0.008   131.802 6.59

my task is to create a column called bin in df , assign index # for each interval of gen.pos for each group. for example, the first bin 1 is assigned to 0~5.964 range of gen.pos , 2 assigned to 5.965 ~ 11.928 ( 5.964*2=11.928 ) ... The final result is like:

mi      chr   gen.pos   bin
m4774   Ch01    0       1
m4775   Ch01    1.701   1
m4663   Ch01    5.519   1
m4777   Ch01    6.5     2
m4779   Ch01    11.067  2
m4780   Ch01    11.234  2
m3933   Ch01    11.449  2
m4782   Ch01    13.986  3
m5534   Ch01    119.277 20
m5536   Ch02    0.036   1
m5550   Ch02    4.26    1

The len data frame output is not necessary. It is just used to describe my question more clearly. Thanks for any helps.

Answer 1

len is important cursor, so I reproduce here for clarity as you did

library(dplyr)
len <- df %>% 
         group_by(chr) %>%
         summarize(min=min(gen.pos), max=max(gen.pos), interval= (max-min)/20) 

Let's say bin width is b=interval , then if x=gen.pos doesn't coincide the endpoints of the intervals, it falls into ceiling((x-min)/b) th interval. So

df %>% 
  group_by(chr) %>% 
  mutate(max   = max(gen.pos), 
         min   = min(gen.pos), 
         width = (max-min)/20, 
         bin1  = ceiling((gen.pos-min)/width),
         bin   = ifelse(gen.pos==min, bin1 + 1, bin1)
         ) 

will produce the desired column with awesome dplyr. (you can ditch the obsolete columns with select command)