我认为您可以在每组NAME中尝试以下操作

PRICE / PRICE[pmax(1, seq_along(PRICE - n)] - 1

例如，对于data.table，

library(data.table)

k <- 1:5
setDT(df)[
  ,
  paste0("ROC", k) := lapply(
    k,
    function(n) PRICE / PRICE[pmax(1, 1:.N - n)] - 1
  ),
  NAME
]

你最终会得到

> df
   NAME PRICE       ROC1       ROC2       ROC3       ROC4       ROC5
1:    A     2  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000
2:    A     3  0.5000000  0.5000000  0.5000000  0.5000000  0.5000000
3:    A     4  0.3333333  1.0000000  1.0000000  1.0000000  1.0000000
4:    A     4  0.0000000  0.3333333  1.0000000  1.0000000  1.0000000
5:    B     7  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000
6:    B     6 -0.1428571 -0.1428571 -0.1428571 -0.1428571 -0.1428571
7:    C     3  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000

数据

> dput(df)
structure(list(NAME = c("A", "A", "A", "A", "B", "B", "C"), PRICE = c(2L,
3L, 4L, 4L, 7L, 6L, 3L)), class = "data.frame", row.names = c(NA,
-7L))

你可以利用跑步者和崩溃来获得更快的速度。
如果没有n的限制，你可以简单地使用collapse::flag()，这将比下面的例子快得多。

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(collapse)
#> collapse 1.9.3, see ?`collapse-package` or ?`collapse-documentation`
#> 
#> Attaching package: 'collapse'
#> The following object is masked from 'package:stats':
#> 
#>     D
library(runner)
library(tidyr)
df <- tibble(NAME = c(rep("A", 4), "B", "B", "C"),
             PRICE = c(2, 3, 4, 4, 7, 6, 3))

# Window vector based on
# k - window size
# n length of vector
# partial window or not?
window_seq <- function (k, n, partial = TRUE){
  stopifnot(length(k) == 1L, length(n) == 1L)
  if (n > .Machine[["integer.max"]]) 
    stop("n must not be greater than .Machine$integer.max")
  n <- as.integer(n)
  k <- as.integer(min(k, n))
  k <- max(k, 0L)
  partial_len <- max(min(k - 1L, n), 0L)
  other_len <- max(0L, n - partial_len)
  if (partial) {
    c(seq_len(partial_len), rep_len(k, other_len))
  }
  else {
    c(rep_len(NA_integer_, partial_len), rep_len(k, other_len))
  }
}

df %>%
  mutate(n1 = 1,
         n2 = 2) %>%
  mutate(g = as.integer(collapse::group(NAME))) %>% # GROUP ID
  mutate(grpn = collapse::GRPN(g)) %>% # GROUP SIZES
  # Using runner we utilise partial windows
  mutate(n1 = window_seq(k = n1[[1]] + 1, 
                         n = grpn[[1]],
                          partial = TRUE),
         n2 = window_seq(k = n2[[1]] + 1, 
                         n = grpn[[1]],
                         partial = TRUE),
         roc1 = (PRICE/runner::lag_run(PRICE, lag = n1 - 1)) - 1,
         roc2 = (PRICE/runner::lag_run(PRICE, lag = n2 - 1)) - 1,
         .by = g)
#> # A tibble: 7 × 8
#>   NAME  PRICE    n1    n2     g  grpn   roc1   roc2
#>   <chr> <dbl> <int> <int> <int> <int>  <dbl>  <dbl>
#> 1 A         2     1     1     1     4  0      0    
#> 2 A         3     2     2     1     4  0.5    0.5  
#> 3 A         4     2     3     1     4  0.333  1    
#> 4 A         4     2     3     1     4  0      0.333
#> 5 B         7     1     1     2     2  0      0    
#> 6 B         6     2     2     2     2 -0.143 -0.143
#> 7 C         3     1     1     3     1  0      0

创建于2023-03-28带有reprex v2.0.2