如何将列联表（计数）转换为 R 中 GLM 的个人

Question

我有这张照片中的信息：

你可以在这里下载：https://drive.google.com/file/d/1pgO51NXtjpVSz-VxQEDNFFuQXVc4jVkt/view?usp=sharing

我想要的是将这些数据转换为个人，

例如

会变成这个

另一个例子

会变成这个

所以，如果我们说n="原始data.frame中所有数字的总和"，即所有个体的数量，最终的输出将是一个6列n行的data.frame。

我想在 R 中执行此操作，但我不知道如何操作。一旦我有了这个，我想做的是应用一个具有家庭二项式和链接=概率的广义线性模型。

现在，这个页面可以解释我尝试做的一些事情：

https://www.datanalytics.com/libro_r/la-funcion-melt-y-datos-en-formato-largo.html

Answer 1

好的...我有一个答案，但是...我想知道是否存在任何概括。就是这样：

library(readxl)
library(dplyr)

# Información original ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
byssinosis <- read_xls(path = "byssinosis.xls",range = "B4:K27",col_names = F)
names(byssinosis) <- c("Employment","Smoking","Sex","Race",
                       "W1y","W1n","W2y","W2n","W3y","W3n")
# View(byssinosis)

# Procesando la información a individuos ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Primero pasamos las columnas a una sola.
datos <- reshape2::melt(byssinosis)
# Separamos estas columnas en las dos características deseadas.
datos <- datos %>%
  mutate(Workplace = ifelse(variable %in% c("W1y", "W1n"),1,
                            ifelse(variable %in% c("W2y", "W2n"),2,3)),
         Byssinosis = ifelse(variable %in% c("W1y", "W2y", "W3y"),"yes","no"))
# Repetimos con base en value.
individuos=rep(seq_len(nrow(datos)),datos$value)
datos <- datos[individuos,]
# Nos quedamos solo las columnas deseadas
datos <- datos %>% select(-c(variable,value))
# View(datos)

# Comprobación ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tabla <-
  table(datos) %>%
  as.data.frame() %>%  
  arrange(Employment, desc(Smoking), desc(Sex), desc(Race), Workplace, desc(Byssinosis))
# View(tabla)

Answer 2

试试这个

library(readxl)
library(dplyr)
library(tidyr)

df <- read_xls("byssinosis.xls", range = cell_rows(c(4L, NA_integer_)), col_names = FALSE)
raw_nms <- read_xls("byssinosis.xls", range = cell_rows(c(1L, 3L)), col_names = FALSE)

names(df) <- with(
  fill(as.data.frame(t(raw_nms)[, -2L]), V1, V2), # replace any missing value in V1 and V2 (i.e. row 1 and 3 in your excel) with the last observation carrired forward
  trimws(paste(V1, if_else(is.na(V2), "", V2))) # collapse these names into a single vector
)

df %>% 
  pivot_longer(contains(" "), names_to = c("Workplace", "byssinosis"), names_pattern = "(\\d+) (.+)") %>% 
  slice(inverse.rle(list(lengths = value, values = seq_along(value)))) %>% 
  select(-value)

输出

# A tibble: 5,419 x 6
   Employment Smoking Sex   Race  Workplace byssinosis
   <chr>      <chr>   <chr> <chr> <chr>     <chr>     
 1 <10        yes     M     W     1         yes       
 2 <10        yes     M     W     1         yes       
 3 <10        yes     M     W     1         yes       
 4 <10        yes     M     W     1         no        
 5 <10        yes     M     W     1         no        
 6 <10        yes     M     W     1         no        
 7 <10        yes     M     W     1         no        
 8 <10        yes     M     W     1         no        
 9 <10        yes     M     W     1         no        
10 <10        yes     M     W     1         no        
# ... with 5,409 more rows