How to create numerical samples based on multiple conditions on multiple vectors?

Question

Given the following data frame:

df <- tibble::tribble(
  ~pass_id, ~km_ini, ~km_fin,
        1L,    0.89,    2.39,
        2L,    1.53,    3.03,
        3L,    21.9,    23.4,
        4L,    23.4,    24.9,
        5L,      24,    25.5,
        6L,    25.9,    27.4,
        7L,    36.7,    38.2,
        8L,    41.4,    42.9,
        9L,    42.1,    43.6,
       10L,    45.5,      47
  )

^{Created on 2020-02-17 by the reprex package (v0.3.0)}

I need a sample of 50 numbers that meet the following criteria for the data frame as a whole, not just for each row of it:

1. >= .750
2. <= 99.450
3. < km_ini - .750
4. > km_fin + .750

What I’ve achieved so far is the easiest part, which are the first two (which I could do directly from the drawing itself with runif - merit 0). The problem is that later I tried to make a enframe and then I tried filter, unsuccessful.

P.S.: I don’t necessarily need the result as a data frame, it can be a vector.

library(tidyverse, verbose = F)

set.seed(42)
sort(runif(100000, 0, 99.450)) %>% 
  enframe(., "ID", "km") %>% 
  filter(km >= .750 & km <= 99.450 - .750)
#> # A tibble: 98,467 x 2
#>       ID    km
#>    <int> <dbl>
#>  1   763 0.750
#>  2   764 0.751
#>  3   765 0.751
#>  4   766 0.753
#>  5   767 0.753
#>  6   768 0.754
#>  7   769 0.754
#>  8   770 0.755
#>  9   771 0.755
#> 10   772 0.757
#> # … with 98,457 more rows

^{Created on 2020-02-17 by the reprex package (v0.3.0)}

EDIT: Trying to visually represent the problem

The final result needs to be a numerical set that evaluates the entire data set, not just each line separately. As an example for the first two lines, see the following representation:

In this way, see that:

• The black line indicates that I cannot have data smaller than . 750.
• The blue line indicates where I can’t have records depending on the coverage area of the km_ini and km_fin (arrows) of line 1 plus an appendix considering the area of + or - 750 (between arrows and dots).
• The red line indicates where I can’t have records depending on the coverage area of the km_ini and km_fin (arrows) of line 2 plus an appendix considering the area of + or - 750 (between arrows and dots).

This way, already face, the random set of data, within the first 4000 meters, could only have numbers from 3030 + 750.

The question, then, is to try to do this programmatically so that all the lines of the data frame are evaluated and the numbers generated are not within all the conditions cited.

Answer 1

I don’t know if I understand the question but if you want n = 50 random numbers described after the last issue of the question perhaps the following code solves the problem.

1. Calculate the minimum and maximum values described above.
2. Create a table with n lines, randomly choosing which lines to replicate.
3. Generates random numbers km within the calculated limits.
4. Discard created auxiliary columns.

If you only need the numbers generated, you can still end with a pipe %>% for select(km).

library(tidyverse, verbose = FALSE)

set.seed(42)

n <- 50

df %>%
  rowwise() %>%
  mutate(Min = pmax(km_ini - 0.750, 0.750),
         Max = pmin(km_fin + 0.750, 99.450)) %>%
  ungroup() %>%
  left_join(tibble(pass_id = sample(.$pass_id, n, TRUE)), by = "pass_id") %>%
  mutate(km = if_else(Min >= 0.750, runif(n(), max(Max), 99.450), runif(n(), 0.750, min(Min)))) %>%
  select(-Min, -Max)
## A tibble: 50 x 4
#   pass_id km_ini km_fin    km
#     <int>  <dbl>  <dbl> <dbl>
# 1       1   0.89   2.39  77.0
# 2       1   0.89   2.39  91.7
# 3       1   0.89   2.39  57.5
# 4       1   0.89   2.39  61.8
# 5       1   0.89   2.39  90.6
# 6       2   1.53   3.03  83.6
# 7       2   1.53   3.03  60.2
# 8       2   1.53   3.03  50.0
# 9       2   1.53   3.03  55.0
#10       2   1.53   3.03  58.9
## … with 40 more rows