从具有分组变量的数据框中随机抽取行的样本。

huangapple
go评论97阅读模式
英文:

Take random sample of rows from dataframe with grouping variables

问题

以下是代码部分的翻译：

  1. 我有一个数据框，结构如下：
  3. dat <- tibble(
  4.   item_type  = rep(1:36, each = 6), 
  5.   condition1 = rep(c("a", "b", "c"), times = 72), 
  6.   condition2 = rep(c("y", "z"), each = 3, times = 36), 
  7. ) %>% 
  8.   unite(unique, item_type, condition1, condition2, sep = "-", remove = TRUE)

看起来像这样：

  1. # 一个 tibble: 216 × 4
  2.    unique item_type condition1 condition2
  3.    <chr>      <int> <chr>      <chr>     
  4.  1 1-a-y          1 a          y         
  5.  2 1-b-y          1 b          y         
  6.  3 1-c-y          1 c          y         
  7.  4 1-a-z          1 a          z         
  8.  5 1-b-z          1 b          z         
  9.  6 1-c-z          1 c          z         
  10.  7 2-a-y          2 a          y         
  11.  8 2-b-y          2 b          y         
  12.  9 2-c-y          2 c          y         
  13. 10 2-a-z          2 a          z

我想随机抽取36行数据。抽样应包括6个 condition1 按 condition2 组合的重复，而不重复 item_type。

使用 slice_sample() 似乎可以得到我想要的子集：

  1. set.seed(1)
  2. dat %>% 
  3.   slice_sample(n = 6, by = c("condition1", "condition2")) %>% 
  4.   count(condition1, condition2)
  1.   condition1 condition2     n
  2. 1 a          y              6
  3. 2 a          z              6
  4. 3 b          y              6
  5. 4 b          z              6
  6. 5 c          y              6
  7. 6 c          z              6

但仔细检查后，我发现 item_type 被重复了。

  1. set.seed(1)
  2. dat %>% 
  3.   slice_sample(n = 6, by = c("condition1", "condition2")) %>% 
  4.   count(item_type) %>% 
  5.   arrange(desc(n))
  1. # 一个 tibble: 22 × 2
  2.    item_type     n
  3.        <int> <int>
  4.  1        10     3
  5.  2        34     3
  6.  3         1     2
  7.  4         6     2
  8.  5         7     2
  9.  6        15     2
  10.  7        20     2
  11.  8        21     2
  12.  9        23     2
  13. 10        25     2
  14. # … 还有更多行

换句话说，我希望从 item_type 中只获得唯一的抽样。是否可能使用 slice_sample() 实现这一点？

编辑
添加第二个示例的数据：

  1. dat <- tibble(
  2.   item_type  = rep(1:36, each = 3), 
  3.   condition1 = rep(c("a", "b"), each = 54), 
  4.   condition2 = rep(c("x", "y", "z"), times = 36), 
  5. ) %>% 
  6.   unite(unique, item_type, condition1, condition2, sep = "-", remove = TRUE)

看起来像这样：

  1. # 一个 tibble: 108 × 4
  2.    unique item_type condition1 condition2
  3.    <chr>      <int> <chr>      <chr>     
  4.  1 1-a-x          1 a          x         
  5.  2 1-a-y          1 a          y         
  6.  3 1-a-z          1 a          z         
  7.  4 2-a-x          2 a          x         
  8.  5 2-a-y          2 a          y         
  9.  6 2-a-z          2 a          z         
  10.  7 3-a-x          3 a          x         
  11.  8 3-a-y          3 a          y         
  12.  9 3-a-z          3 a          z         
  13. 10 4-a-x          4 a          x

尝试进行抽样：

  1. inner_join(
  2.   dat, 
  3.   distinct(dat, condition1, condition2) %>% 
  4.     uncount(n()) %>% 
  5.     mutate(item_type = sample(n()))
  6. )

这将生成一个长度为20的数据框，具有以下特点：

  1.   condition1 condition2     n
  2. 1 a          x              4
  3. 2 a          y              4
  4. 3 a          z              4
  5. 4 b          x              3
  6. 5 b          y              4
  7. 6 b          z              5
英文:

I have a dataframe with the following structure:

  1. dat &lt;- tibble(
  2.   item_type  = rep(1:36, each = 6), 
  3.   condition1 = rep(c(&quot;a&quot;, &quot;b&quot;, &quot;c&quot;), times = 72), 
  4.   condition2 = rep(c(&quot;y&quot;, &quot;z&quot;), each = 3, times = 36), 
  5. ) %&gt;% 
  6.   unite(unique, item_type, condition1, condition2, sep = &quot;-&quot;, remove = F)

which looks like this:

  1. # A tibble: 216 &#215; 4
  2.    unique item_type condition1 condition2
  3.    &lt;chr&gt;      &lt;int&gt; &lt;chr&gt;      &lt;chr&gt;     
  4.  1 1-a-y          1 a          y         
  5.  2 1-b-y          1 b          y         
  6.  3 1-c-y          1 c          y         
  7.  4 1-a-z          1 a          z         
  8.  5 1-b-z          1 b          z         
  9.  6 1-c-z          1 c          z         
  10.  7 2-a-y          2 a          y         
  11.  8 2-b-y          2 b          y         
  12.  9 2-c-y          2 c          y         
  13. 10 2-a-z          2 a          z

I would like to take a random sample of 36 rows. The sample should include 6 repetitions of the condition1 by condition2 combinations without repeating item_type.

Using slice_sample() it seems I can get the subset I want...

  1. set.seed(1)
  2. dat %&gt;% 
  3.   slice_sample(n = 6, by = c(&quot;condition1&quot;, &quot;condition2&quot;)) %&gt;% 
  4.   count(condition1, condition2)
  1.   condition1 condition2     n
  2.   &lt;chr&gt;      &lt;chr&gt;      &lt;int&gt;
  3. 1 a          y              6
  4. 2 a          z              6
  5. 3 b          y              6
  6. 4 b          z              6
  7. 5 c          y              6
  8. 6 c          z              6

But on closer inspection I see that item_type is repeated.

  1. set.seed(1)
  2. dat %&gt;% 
  3.   slice_sample(n = 6, by = c(&quot;condition1&quot;, &quot;condition2&quot;)) %&gt;% 
  4.   count(item_type) %&gt;% 
  5.   arrange(desc(n))
  1. # A tibble: 22 &#215; 2
  2.    item_type     n
  3.        &lt;int&gt; &lt;int&gt;
  4.  1        10     3
  5.  2        34     3
  6.  3         1     2
  7.  4         6     2
  8.  5         7     2
  9.  6        15     2
  10.  7        20     2
  11.  8        21     2
  12.  9        23     2
  13. 10        25     2
  14. # … with 12 more rows

In other words, I would like only unique pulls overall from item_type.
Is it possible to get slice_sample() to do this?

EDIT
Adding second toy data example.

  1. dat &lt;- tibble(
  2.   item_type  = rep(1:36, each = 3), 
  3.   condition1 = rep(c(&quot;a&quot;, &quot;b&quot;), each = 54), 
  4.   condition2 = rep(c(&quot;x&quot;, &quot;y&quot;, &quot;z&quot;), times = 36), 
  5. ) %&gt;% 
  6.   unite(unique, item_type, condition1, condition2, sep = &quot;-&quot;, remove = F)

Which looks like this:

  1. # A tibble: 108 &#215; 4
  2.    unique item_type condition1 condition2
  3.    &lt;chr&gt;      &lt;int&gt; &lt;chr&gt;      &lt;chr&gt;     
  4.  1 1-a-x          1 a          x         
  5.  2 1-a-y          1 a          y         
  6.  3 1-a-z          1 a          z         
  7.  4 2-a-x          2 a          x         
  8.  5 2-a-y          2 a          y         
  9.  6 2-a-z          2 a          z         
  10.  7 3-a-x          3 a          x         
  11.  8 3-a-y          3 a          y         
  12.  9 3-a-z          3 a          z         
  13. 10 4-a-x          4 a          x

Attempt to sample:

  1. inner_join(
  2.   dat, 
  3.   distinct(dat,condition1, condition2) %&gt;% 
  4.     uncount(n()) %&gt;% 
  5.     mutate(item_type = sample(n()))
  6. )

Which produces a dataframe of length 20 with the following characteristics:

  1.   condition1 condition2     n
  2.   &lt;chr&gt;      &lt;chr&gt;      &lt;int&gt;
  3. 1 a          x              4
  4. 2 a          y              4
  5. 3 a          z              4
  6. 4 b          x              3
  7. 5 b          y              4
  8. 6 b          z              5

答案1

得分: 2

以下是您要翻译的代码部分：

  1. You could do this:

inner_join(
dat,
distinct(dat,condition1, condition2) %>%
uncount(n()) %>%
mutate(item_type=sample(n())),
)

  2. Output:

A tibble: 36 × 4

unique item_type condition1 condition2
<chr> <int> <chr> <chr>
1 1-b-z 1 b z
2 2-a-z 2 a z
3 3-c-y 3 c y
4 4-c-z 4 c z
5 5-b-z 5 b z
6 6-a-y 6 a y
7 7-c-y 7 c y
8 8-a-y 8 a y
9 9-a-y 9 a y
10 10-c-z 10 c z

… with 26 more rows

  2. On the second dataset, you need to get the min/max range to sample:
  3. ```R
  4. inner_join(
  5.   dat, 
  6.   distinct(dat,condition1, condition2) %&gt;% 
  7.     uncount(n()) %&gt;% 
  8.     inner_join(dat %&gt;% group_by(condition1, condition2) %&gt;% summarize(imin = min(item_type), imax=max(item_type), .groups=&quot;drop&quot;)) %&gt;% 
  9.     group_by(condition1) %&gt;% 
  10.     mutate(item_type = sample(imin[1]:imax[1],size = n())) %&gt;% 
  11.     ungroup() %&gt;% 
  12.     select(-c(imin:imax))
  13. )

Output:

  1. # A tibble: 36 &#215; 4
  2.    unique item_type condition1 condition2
  3.    &lt;chr&gt;      &lt;int&gt; &lt;chr&gt;      &lt;chr&gt;     
  4.  1 1-a-y          1 a          y         
  5.  2 2-a-z          2 a          z         
  6.  3 3-a-z          3 a          z         
  7.  4 4-a-y          4 a          y         
  8.  5 5-a-z          5 a          z         
  9.  6 6-a-y          6 a          y         
  10.  7 7-a-x          7 a          x         
  11.  8 8-a-z          8 a          z         
  12.  9 9-a-y          9 a          y         
  13. 10 10-a-z        10 a          z         
  14. # … with 26 more rows
英文:

You could do this:

  1. inner_join(
  2.   dat, 
  3.   distinct(dat,condition1, condition2) %&gt;% 
  4.     uncount(n()) %&gt;% 
  5.     mutate(item_type=sample(n())),
  6. )

Output:

  1. # A tibble: 36 &#215; 4
  2.    unique item_type condition1 condition2
  3.    &lt;chr&gt;      &lt;int&gt; &lt;chr&gt;      &lt;chr&gt;     
  4.  1 1-b-z          1 b          z         
  5.  2 2-a-z          2 a          z         
  6.  3 3-c-y          3 c          y         
  7.  4 4-c-z          4 c          z         
  8.  5 5-b-z          5 b          z         
  9.  6 6-a-y          6 a          y         
  10.  7 7-c-y          7 c          y         
  11.  8 8-a-y          8 a          y         
  12.  9 9-a-y          9 a          y         
  13. 10 10-c-z        10 c          z         
  14. # … with 26 more rows

On the second dataset, you need to get the min/max range to sample:

  1. inner_join(
  2.   dat, 
  3.   distinct(dat,condition1, condition2) %&gt;% 
  4.     uncount(n()) %&gt;% 
  5.     inner_join(dat %&gt;% group_by(condition1, condition2) %&gt;% summarize(imin = min(item_type), imax=max(item_type), .groups=&quot;drop&quot;)) %&gt;% 
  6.     group_by(condition1) %&gt;% 
  7.     mutate(item_type = sample(imin[1]:imax[1],size = n())) %&gt;% 
  8.     ungroup() %&gt;% 
  9.     select(-c(imin:imax))
  10. )

Output:

  1. # A tibble: 36 &#215; 4
  2.    unique item_type condition1 condition2
  3.    &lt;chr&gt;      &lt;int&gt; &lt;chr&gt;      &lt;chr&gt;     
  4.  1 1-a-y          1 a          y         
  5.  2 2-a-z          2 a          z         
  6.  3 3-a-z          3 a          z         
  7.  4 4-a-y          4 a          y         
  8.  5 5-a-z          5 a          z         
  9.  6 6-a-y          6 a          y         
  10.  7 7-a-x          7 a          x         
  11.  8 8-a-z          8 a          z         
  12.  9 9-a-y          9 a          y         
  13. 10 10-a-z        10 a          z         
  14. # … with 26 more rows

答案2

得分: 1

以下是您要求的代码翻译：

Try

  1. library(nplyr)
  2. library(dplyr)
  3. library(tidyr)
  4. dat %>%
  5.   nest(data = -item_type) %>%
  6.   nest_slice_sample(data, n = 1) %>%
  7.   unnest(data)

-output

  1. # A tibble: 36 × 4
  2.    item_type unique condition1 condition2
  3.        <int> <chr>  <chr>      <chr>     
  4.  1         1 1-c-z  c          z         
  5.  2         2 2-b-z  b          z         
  6.  3         3 3-b-y  b          y         
  7.  4         4 4-c-y  c          y         
  8.  5         5 5-c-z  c          z         
  9.  6         6 6-b-z  b          z         
  10.  7         7 7-a-z  a          z         
  11.  8         8 8-c-z  c          z         
  12.  9         9 9-b-y  b          y         
  13. 10        10 10-a-y a          y         
  14. # … with 26 more rows

Or perhaps we need

  1. lst1 <- split(dat, dat[c("condition1", "condition2")], drop = TRUE)
  2. lst2 <- vector('list', length(lst1))
  3. item_type_rm <- numeric(0)
  4. for(i in seq_along(lst1))
  5. {
  6.   tmp <- lst1[[i]]
  7.   tmp1 <- tmp %>%
  8.      filter(!item_type %in% item_type_rm) %>%
  9.      slice_sample(n = 6)
  10.    item_type_rm <- c(item_type_rm, tmp1$item_type)
  11.    lst2[[i]] <- tmp1
  12. }
  13. out <- bind_rows(lst2)
  14. out
  15. # A tibble: 36 × 4
  16.    unique item_type condition1 condition2
  17.    <chr>      <int> <chr>      <chr>     
  18.  1 17-a-x        17 a          x         
  19.  2 5-a-x          5 a          x         
  20.  3 9-a-x          9 a          x         
  21.  4 2-a-x          2 a          x         
  22.  5 7-a-x          7 a          x         
  23.  6 3-a-x          3 a          x         
  24.  7 31-b-x        31 b          x         
  25.  8 27-b-x        27 b          x         
  26.  9 36-b-x        36 b          x         
  27. 10 19-b-x        19 b          x         
  28. # … with 26 more rows
  29. > out %>% count(item_type) %>% pull(n)
  30.  [1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

请注意，这里只翻译了代码部分，没有包括注释和输出。

英文:

Try

  1. library(nplyr)
  2. library(dplyr)
  3. library(tidyr)
  4. dat %&gt;% 
  5.   nest(data = -item_type) %&gt;%
  6.   nest_slice_sample(data, n = 1) %&gt;%
  7.   unnest(data)

-output

  1. # A tibble: 36 &#215; 4
  2.    item_type unique condition1 condition2
  3.        &lt;int&gt; &lt;chr&gt;  &lt;chr&gt;      &lt;chr&gt;     
  4.  1         1 1-c-z  c          z         
  5.  2         2 2-b-z  b          z         
  6.  3         3 3-b-y  b          y         
  7.  4         4 4-c-y  c          y         
  8.  5         5 5-c-z  c          z         
  9.  6         6 6-b-z  b          z         
  10.  7         7 7-a-z  a          z         
  11.  8         8 8-c-z  c          z         
  12.  9         9 9-b-y  b          y         
  13. 10        10 10-a-y a          y         
  14. # … with 26 more rows

Or perhaps we need

  1. lst1 &lt;- split(dat, dat[c(&quot;condition1&quot;, &quot;condition2&quot;)], drop = TRUE)
  2. lst2 &lt;- vector(&#39;list&#39;, length(lst1))
  3. item_type_rm &lt;- numeric(0)
  4. for(i in seq_along(lst1))
  5. {
  6. tmp &lt;- lst1[[i]]
  7. tmp1 &lt;- tmp %&gt;% 
  8.    filter(!item_type %in% item_type_rm) %&gt;%
  9.    slice_sample(n = 6)
  10.  item_type_rm &lt;- c(item_type_rm, tmp1$item_type)
  11.  lst2[[i]] &lt;- tmp1
  12.  
  14. }
  16. out &lt;- bind_rows(lst2)
  17. out
  18. # A tibble: 36 &#215; 4
  19.    unique item_type condition1 condition2
  20.    &lt;chr&gt;      &lt;int&gt; &lt;chr&gt;      &lt;chr&gt;     
  21.  1 17-a-x        17 a          x         
  22.  2 5-a-x          5 a          x         
  23.  3 9-a-x          9 a          x         
  24.  4 2-a-x          2 a          x         
  25.  5 7-a-x          7 a          x         
  26.  6 3-a-x          3 a          x         
  27.  7 31-b-x        31 b          x         
  28.  8 27-b-x        27 b          x         
  29.  9 36-b-x        36 b          x         
  30. 10 19-b-x        19 b          x         
  31. # … with 26 more rows
  32. &gt; out %&gt;% count(item_type) %&gt;% pull(n)
  33.  [1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

huangapple
  • 本文由 发表于 2023年2月27日 07:22:54
  • 转载请务必保留本文链接：https://go.coder-hub.com/75575621.html
  • r
  • random
  • tidyverse
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定