Error in UseMethod("select") : no applicable method for 'select' applied to an object of class "character"

Question

This question is very similar to this one Make coefficient for all dates/categories, what is different is a few things in the return_coef function. You will see that I can generate the coefficients for each day/category, but when I ask to do it for everyone at once, I get the following error:

Error in UseMethod("select") : 
  no applicable method for 'select' applied to an object of class "character" 

Executable code below:

library(dplyr)
library(tidyverse)
library(lubridate)


  df1 <- structure(
    list(date1= c("2021-06-26","2021-06-26","2021-06-26","2021-06-26"),
         date2 = c("2021-06-27","2021-07-01","2021-07-02","2021-07-03"),
         Category = c("ABC","ABC","ABC","ABC"),
         Week= c("Saturday","Wednesday","Thurday","Saturday"),
         DR1 = c(5,4,1,1),
         DRM01 = c(8,4,1,0), DRM02= c(7,4,2,0),DRM03= c(6,9,5,0),
         DRM04 = c(5,5,4,0),DRM05 = c(5,5,4,0),DRM06 = c(7,5,4,0),DRM07 = c(2,5,4,0),DRM08 = c(2,5,4,0)),
    class = "data.frame", row.names = c(NA, -4L))


  return_coef <- function(df1, dmda, CategoryChosse, var1, var2, gnum=0, graf=1) {
  
  x<-df1 %>% select(starts_with("DRM0"))
  
  x<-cbind(df1, setNames(df1$DR1 - x, paste0(names(x), "_PV")))
  PV<-select(x, date2,Week, Category, DR1, ends_with("PV"))
  
  med<-PV %>%
    group_by(Category,Week) %>%
    dplyr::summarize(dplyr::across(ends_with("PV"), median))
  
  SPV<-df1%>%
    inner_join(med, by = c('Category', 'Week')) %>%
    mutate(across(matches("^DRM0\\d+$"), ~.x + 
                    get(paste0(cur_column(), '_PV')),
                  .names = '{col}_{col}_PV')) %>%
    select(date1:Category, DRM01_DRM01_PV:last_col())
  
  SPV<-data.frame(SPV)
  
  mat1 <- df1 %>%
    dplyr::filter(date2 == dmda, Category == CategoryChosse) %>%
    select(starts_with("DRM0")) %>%
    pivot_longer(cols = everything()) %>%
    arrange(desc(row_number())) %>%
    mutate(cs = cumsum(value)) %>%
    dplyr::filter(cs == 0) %>%
    pull(name)
  
  (dropnames <- paste0(mat1,"_",mat1, "_PV"))
  
  SPV <- SPV %>%
    filter(date2 == dmda, Category == CategoryChosse) %>%
    select(-any_of(dropnames))
  
  if(length(grep("DRM0", names(SPV))) == 0) {
    SPV[head(mat1,10)] <- NA_real_
  }
  
  datas <-SPV %>%
    dplyr::filter(date2 == ymd(dmda)) %>%
    group_by(Category) %>%
    dplyr::summarize(dplyr::across(starts_with("DRM0"), sum)) %>%
    pivot_longer(cols= -Category, names_pattern = "DRM0(.+)", values_to = "val") %>%
    mutate(name = readr::parse_number(name))
  colnames(datas)[-1]<-c(var1,var2)
  datas$days <- datas[[as.name(var1)]]
  datas$numbers <- datas[[as.name(var2)]]
  
  datas <- datas %>% 
    group_by(Category) %>% 
    slice((as.Date(dmda) - min(as.Date(df1$date1) [
      df1$Category == first(Category)])):max(days)+1) %>%
    ungroup
  
  m<-df1 %>%
    group_by(Category,Week) %>%
    dplyr::summarize(dplyr::across(starts_with("DR1"), mean))
  
  m<-subset(m, Week == df1$Week[match(ymd(dmda), ymd(df1$date2))] & Category == CategoryChosse)$DR1
  
  if (nrow(datas)<=2){
    val<-as.numeric(m)
  }
  
  else{
    mod <- nls(numbers ~ b1*days^2+b2,start = list(b1 = 0,b2 = 0),data = datas, algorithm = "port")
    coef<-coef(mod)[2]
    val<-as.numeric(coef(mod)[2])
  }
  
  
  return(val)
  
}

All<-cbind(df1 %>% select(date2, Category), coef = mapply(return_coef, df1$date2, df1$Category))

Error in UseMethod("select") : 
  no applicable method for 'select' applied to an object of class "character"

If I want to know the coefficient for each one separately, I can do it.

return_coef(df1, "2021-06-27","ABC", var1=0,var2=1)
[1] 6.539702
return_coef(df1, "2021-07-01","ABC", var1=0,var2=1)
[1] 4
return_coef(df1, "2021-07-02","ABC", var1=0,var2=1)
[1] 1
return_coef(df1, "2021-07-03","ABC", var1=0,var2=1)
[1] 3

Answer 1

Two problems:

• The first argument for your return_coef function is a data.frame named df1, yet you are calling it with df1$date2 (a string). I think you should instead start with

  mapply(return_coef, list(df1), df1$date2, df1$Category)
  

  (though this does error currently, see the next bullet).

  The list(df1) in this case means that the whole df1 will be passed as the first argument for each of the pairs from df1$date2 and df1$Category.

• It now fails with argument "var1" is missing, with no default, but I suspect you were working towards that. I'll choose a couple of random names and ... something happens.

Ultimately, the function is fine as-is, just change your mapply use as:

mapply(return_coef, list(df1), df1$date2, df1$Category, var1 = "a1", var2 = "a2")
# [1] 6.539702 4.000000 1.000000 3.000000

Because both var1 and var2 are length-1, they are recycled for all calls to return_coef (as their named arguments).

Since you're using dplyr, this can be neatly put into a pipe a little more directly than using cbind(...):

library(dplyr)
  df1 %>%
    transmute(
      date2, Category,
      coef = mapply(return_coef, list(cur_data()), date2, Category, var1 = "a1", var2 = "a2")
    )
#        date2 Category     coef
# 1 2021-06-27      ABC 6.539702
# 2 2021-07-01      ABC 4.000000
# 3 2021-07-02      ABC 1.000000
# 4 2021-07-03      ABC 3.000000

I use transmute instead of a preceding select(date2, Category) because the function needs variables present in the whole frame. I could easily have done mutate(coef=..) %>% select(date2, Category, coef) as well.