在级别中置换列，对2列进行测试，然后保存pvalues

如何解决在级别中置换列，对2列进行测试，然后保存pvalues

我有一个数据框

> dput(df)
structure(list(id = c(1,2,3,4,1,4),level = structure(c(1L,1L,2L,2L),.Label = c("g01","g02"),class = "factor"),m_col = c(1,11,22,33,44),u_col = c(11,12,13,14,21,23,24),group = c(0,1
    )),row.names = c(NA,-8L),class = "data.frame")

看起来像这样

  id level m_col u_col group
1  1   g01     1    11     0
2  2   g01     2    12     0
3  3   g01     3    13     1
4  4   g01     4    14     1
5  1   g02    11    21     0
6  2   g02    22    22     0
7  3   g02    33    23     1
8  4   g02    44    24     1

我想在每个“级别”上执行二项式加权测试（本质上，我需要比较每个id的u_col和m_col）...因此，使用tidyverse和broom我可以以下：

res <- df %>% 
  group_by(level) %>% 
  do(tidy(glm(cbind(.$m_col,.$u_col) ~ .$group,family="binomial"))) %>%
  filter(term == ".$group")

每个级别为我提供了一些p值：

> res
# A tibble: 2 x 6
# Groups:   level [2]
  level term    estimate std.error statistic p.value
  <fct> <chr>      <dbl>     <dbl>     <dbl>   <dbl>
1 g01   .$group    0.687     0.746     0.921  0.357 
2 g02   .$group    0.758     0.296     2.56   0.0105

然后我可以问多少p

length(which(res$p.value < 0.05)

我现在想对数据进行置换，重复二项式检验，询问有多少p

但是，排列需要重新排列每个“级别”中的“组”列。我正在努力寻找一种方法来做到这一点，例如，一个排列看起来像这样

  id level m_col u_col group
1  1   g01     1    11     1
2  2   g01     2    12     0
3  3   g01     3    13     1
4  4   g01     4    14     0
5  1   g02    11    21     1
6  2   g02    22    22     0
7  3   g02    33    23     1
8  4   g02    44    24     0

一秒钟会变成

  id level m_col u_col group
1  1   g01     1    11     0
2  2   g01     2    12     1
3  3   g01     3    13     1
4  4   g01     4    14     0
5  1   g02    11    21     0
6  2   g02    22    22     1
7  3   g02    33    23     1
8  4   g02    44    24     0

等

让测试依赖于2列会限制洗牌选项，这让我很沮丧。我将不胜感激。

解决方法

如果您想要一个数据框，可以尝试以下操作：

<input  name="study[field]">
<input  name="study[branch]">
<input  name="study[grade_id]">
<input  name="study[institution_id]">
<input  name="study[institution_education]">
<input  name="study[gpa]">
<input  name="study[nation_id]">
<input  name="study[province_id]">
<input  name="study[town_id]">
<input  name="study[branch]">
<input  name="study[province_name]">
<input  name="study[town_name]">
<input  name="study[entrance]">
<input  name="study[graduate]">
<input  name="study[currently_studying]">

，如果需要向量：

    foreach ($request->field as $key => $value) {
        Educational::firstOrCreate(['user_id' => auth()->id()],[
            'grade_id' => $request->grade_id[$key],'field' => $request->field[$key],'institution_id' => $request->institution_id[$key],'branch' => $request->branch[$key],'institution_education' => $request->institution_education[$key],'gpa' => $request->gpa[$key],'nation_id' => $request->nation_id[$key],'province_id' => $request->province_id[$key],'town_id' => $request->town_id[$key],'province_name' => $request->province_name[$key],'town_name' => $request->town_name[$key],'entrance' => $request->entrance[$key],'graduate' => $request->graduate[$key],'currently_studying' => $request->has("currently_studying.$key"),]);
    }

,

您可以编写一个函数：

library(dplyr)
library(broom)

apply_fun <- function(data) {
  sum(subset(tidy(glm(cbind(m_col,u_col)~group,data,family="binomial")),term == 'group')$p.value < 0.05)
}

，然后使用replicate重复它。

result <- replicate(100,df %>%
  group_by(level) %>%
  mutate(group = sample(group)) %>%
  summarise(value = apply_fun(cur_data())),simplify = FALSE)
result