R大数据框基于网格的随机子样本

Question

我有一个非常大的数据框（7000 列和 14000 个）观察值。它们实际上是对图像每个像素的灰度观察。 x 轴上 7000 像素的观察值和 y 轴上 140000 像素的观察值。 我正在寻找一种方法来执行以下操作：

• 将数据框划分为 1000 x 1000 像素（或 1000 列和 1000 行）的网格。在这种情况下，它将是一个 7 x 14 的网格
• 从每个网格中随机选择 1、2 或更多像素
• 将返回的值连同其 x 和 y 坐标以及它的来源网格（如果可能）存储在新的数据框中，但这意味着也要对每个网格进行编号。

任何关于我如何做到这一点的想法将不胜感激

Answer 1

如果你想有块矩阵，你可以试试下面的代码

blks <- t(
  sapply(
    split(
      df,
      ceiling(seq(nrow(df)) / 1000)
    ),
    function(x) {
      Map(
        as.matrix,
        split.default(x, ceiling(seq_along(x) / 1000))
      )
    }
  )
)

你会看到

> blks
  1               2               3               4
1 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
2 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
3 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
4 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
5 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
6 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
7 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
  5               6               7               8
1 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
2 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
3 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
4 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
5 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
6 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
7 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
  9               10              11              12
1 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
2 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
3 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
4 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
5 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
6 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
7 Numeric,1000000 Numeric,1000000 Numeric,1000000 Numeric,1000000
  13              14
1 Numeric,1000000 Numeric,1000000
2 Numeric,1000000 Numeric,1000000
3 Numeric,1000000 Numeric,1000000
4 Numeric,1000000 Numeric,1000000
5 Numeric,1000000 Numeric,1000000
6 Numeric,1000000 Numeric,1000000
7 Numeric,1000000 Numeric,1000000

> dim(blks[[2,3]])
[1] 1000 1000

> str(blks[[2,3]])
 num [1:1000, 1:1000] 0.909 0.833 0.347 0.837 0.58 ...
 - attr(*, "dimnames")=List of 2
  ..$ : chr [1:1000] "1001" "1002" "1003" "1004" ...
  ..$ : chr [1:1000] "X2001" "X2002" "X2003" "X2004" ...

数据

set.seed(123)
df <- data.frame(matrix(runif(7000 * 14000), ncol = 14000))

Answer 2

您可以通过选择所需的随机观察结果来进行下采样，然后对数据框进行子集化。例如，下面的代码会在每个块中生成一个随机位置，然后对选择的点进行子集化。

library(tidyverse)
set.seed(42)
df <- data.frame(matrix(runif(7000 * 14000), ncol = 7000, nrow = 14000))
h <- 14
w <- 7
j <- 1000
x_places <- sample.int(j, w, replace = TRUE) + seq(0, (w-1) * j, by = j)
y_places <- sample.int(j, h, replace = TRUE) + seq(0, (h-1) * j, by = j)
new_df <- df[y_places, x_places]

如果您想要更精细或更粗糙的网格，您可以相应地更改 h、j 和 k。此外，如果您为原始数据框的行和列命名，则默认情况下您将拥有这些位置。

Answer 3

遵循这个策略

#create a sample dataframe
set.seed(123)
df <- data.frame(matrix(runif(7000*14000), 14000))

#Step-1: create a blank output df say `df2`
df2 <- as.data.frame(matrix(rep(NA, 1000*1000), 1000))
#step-2: for loop to store sampled values in output `df2`
for(i in 0:999){
  for(j in 0:999){
    df2[i+1, j+1] <- sample(matrix(as.matrix(df[0:13999 %/% 14 == i, 0:6999 %/% 7 == j]),1),1)
  }
}

检查它的尺寸

> dim(df2)
[1] 1000 1000

检查其随机元素以查看该循环是否有效

> df2[5,45]
[1] 0.1724635

Answer 4

我确信比我更好的人有一个更优雅的解决方案，但这就是我最终所做的，在一个 7000 x 140000 的矩阵中随机抽样每个 1000 x 1000 网格。它冗长但有效

设置随机数（1 和 1000）的最小值和最大值并逐行工作以随机选择行值和列值并将该数字从矩阵中提取出来。

第一行 (7000/1000) 执行此操作七次。 然后将 1000 添加到第二行的最小值和最大值 每次提取这些值时，它们都会存储到“g”并使用 rbind 附加到 df_new。

重复直到我到达第 14 行，他们整理了一些数据并重命名。

最后是一个包含 98 个值的数据框，以及它们来自的网格编号及其在矩阵中的 x 和 y 坐标

对于 7000 x 14000 的矩阵和 1000 x 1000 的网格不是很灵活和固定。但我不需要改变这个.....但是！！！

我的愿望清单将是一个让我设置的功能..

• 网格大小（x 和 y），所以我可以有矩形网格
• 从每个网格中提取的样本数（1、2、3 或更多）

感谢大家的意见。非常感谢。

# MAKE A DUMMY MATRIX
set.seed(123)
df <- data.frame(matrix(runif(7000*14000), 14000))

# PICK 1 RANDOM VALUE FROM EACH 1000 BY 1000 GRID
# STARTING WITH THE TOP ROW (GRIDS 1 TO 7)
# THEN THE NEXT ROW (GRIDS 8 TO 9) ETC ETC 
# UNITL GRID 98

set.seed(42)

# 1st row of grids
min=1
max=1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
df_new = c(ran_c, ran_r, df[ran_r,ran_c])

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 2nd row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 3rd row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 4th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 5th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 6th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 7th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 8th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 9th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 10th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 11th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 12th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 13th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

# 14th row of grids
min = min + 1000
max = max +1000

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1,1000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,1001,2000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,2001,3000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,3001,4000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,4001,5000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,5001,6000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)

ran_r = round(runif(1,min,max),0)
ran_c = round(runif(1,6001,7000),0)
g = c(ran_c, ran_r, df[ran_r,ran_c])
df_new = rbind(df_new, g)


# SOME DATA TIDYING
library(dplyr)
df_new = as.data.frame(df_new) # convert to data frame
df_new = df_new %>% mutate(grid = 1:n()) # add a sequential column named grid
df_new = df_new %>%
  relocate(grid)
rownames(df_new)<-c(1:nrow(df_new)) # add row names 1 to 98 or nrow(df_new)
df_new = df_new %>%
  rename(
    x = 2,
    y = 3,
    value = 4) # rename columns

# Clear data no longer needed to prepare for next run
rm(g)
rm(max)
rm(min)
rm(ran_c)
rm(ran_r)