dplyr function group_by error

Question

I have problems with a function of library dplyr. I want to group a dataframe by various values ("group_by"). Some of these values are fix (always the same), and some are introduced through a vector. This vector would have variable dimensions. When the data frame will be grouped, I want to apply the function "mutate".

I have tryed to do it by different ways. The first is copied below, and includes a loop that goes over the vector campToAgregate (where are located the values needed to group the dataframe):

campToAgregate = c("via","nomDem")

dadesCom <- dades 

for(i in 1:length(campToAgregate)){
  if(i==1){
  dadesCom1 <- dadesCom %>% dplyr::group_by(dadesCom[,which(names(dadesCom) == campToAgregate[i])], dat, add=TRUE) %>%
               dplyr::mutate(vel1 = round(weighted.mean(vel, longPk, na.rm = TRUE), 0))
  dadesCom1 <- dadesCom1[,-(ncol(dadesCom1)-1)]
  }else{
  dadesCom2 <- dadesCom1 %>% dplyr::group_by(dadesCom1[,which(names(dadesCom1) == campToAgregate[i])], add=TRUE) %>%
               dplyr::mutate(vel1 = round(weighted.mean(vel, longPk, na.rm = TRUE), 0))
  }
  }

dades is the dataframe, and it contains a lot of values, including the values that are mentioned in the function above: "vel" and "longPk".

When I run this code, the following error appears in the console:

Error in mutate_impl(.data, dots) : not compatible with STRSXP

And I don't know how to solve it...

I have also tryed to do it by a different way:

for(i in 1:length(campToAgregate)){
  if(i==1){
    dadesCom <- dadesCom %>% dplyr::group_by(dadesCom[,which(names(dadesCom) == campToAgregate[i])], dat, add=TRUE)
  }else{
    dadesCom <- dadesCom %>% dplyr::group_by(dadesCom1[,which(names(dadesCom1) == campToAgregate[i])], add=TRUE)
  }
}

dadesCom <- dadesCom %>% dplyr::mutate(vel = round(weighted.mean(vel, longPk, na.rm = TRUE), 0))

But in this case the function group_by doesn't work. The mutate function works, but it's applied to the dataframe without group.

Does anybody know what kind of mistakes I'm doing in the code? Thank you.

Answer 1

This can be accomplished using tidy evaluation semantics. Here is an example using mtcars as no sample data was provided:

library(dplyr)

ag <- c(quo(cyl), quo(gear))

lapply(ag, function(x) mtcars %>%
         group_by(!!x) %>%
         mutate(vel1 = round(weighted.mean(hp, wt, na.rm = TRUE), 0)))

Depending on the desired output summarise might be a better suited function as it will only display one row for each group

lapply(ag, function(x) mtcars %>%
         group_by(!!x) %>%
         summarise(vel1 = round(weighted.mean(hp, wt, na.rm = TRUE), 0)))

[[1]]
# A tibble: 3 x 2
    cyl  vel1
  <dbl> <dbl>
1     4    83
2     6   122
3     8   209

[[2]]
# A tibble: 3 x 2
   gear  vel1
  <dbl> <dbl>
1     3   182
2     4    94
3     5   219

Answer 2

I was able to reproduce the error. Testing the code piecemeal, we find that

dadesCom2 <- dadesCom1 %>%
               dplyr::group_by(dadesCom1[,which(names(dadesCom1) == campToAgregate[i])], add=TRUE) %>%
               dplyr::mutate(vel1 = round(weighted.mean(vel, longPk, na.rm = TRUE), 0))

produces this error

Error in grouped_df_impl(data, unname(vars), drop) : Column dadesCom1[, which(names(dadesCom1) == i)] can't be used as a grouping variable because it's a tbl_df/tbl/data.frame

Simply add

dadesCom1 <- as.data.frame(dadesCom1)

to the end of your first statement.

I'd also suggest using library(dplyr) and removing your in-line calls