Double Loops in R: Use .name_repair to specify repair?-CodePudding

The

When extracting the group, you selected a dataset row, not a unique value.

# this selects jth row, then looks for unique values groups_j = unique(example$group[j]) # you need to get the unique values, then iterate group_j = unique(example$group)[j] # the j goes outside the call for unique()

In this code, you wrote file and file$fsa. I'm assuming this is equivalent to example and example$group, because I don't have whatever is in file.

All of these lines of code do the same thing. Keep in mind that group is in the data frame, but it's an object in the environment on its own, as well.

my_data_i = example[which(example$group == group_j), ] # this would work my_data_i <- filter(example, groups == group_j) # this would work my_data_i <- example[group == group_j, ] # this would work my_data_i <- example[example$group == group_j, ] # this would work

If you iterated over groups instead of indices, you can skip the creation of group_j which is the only time you used j.

for(j in unique(example$group)) { my_data_i <- example[example$group == j, ] }

When you sent the results using i and i alone, you would overwrite the data with each iteration over j.

The first group iteration can go in results[[i]], but the next group, can either be bound to that data or placed in a list within a list.

For example:

results[[1]] <- group 1, method 1 results[[1]] <- rbind(results[[1]], [group 2, method 1]) # or results[[1]][[1]] <- group 1, method 1 results[[1]][[2]] <- group 2, method 1

Considering those two options for the lists (above), the first will allow your remaining code (creation of final, averages, etc.) to work without any changes. However, if you use the second option (above), that code will require modification.

If you leave for(j with 1:length this will work:

if(j < 2) { results[[i]] <- final_i } else { results[[i]] <- rbind(results[[i]], final_i) }

If you use for(j in unique(example$group)), you could use this:

if(isTRUE(j == unique(example$group)[1])) { # isTRUE() to avoid null errors results[[i]] <- final_i } else { results[[i]] <- rbind(results[[i]], final_i) }

Your nested for statements all in one chunk.

results = list() for (i in 1:length(method)) { # bracket missing here; it was in the wrong place for (j in 1:length(unique(example$group))) { # missing a parentheses here # { # this needs to be after each for statement # groups_j = unique(example$group[j]) # you have selected the jth row, not the jth unique group_j = unique(example$group)[j] # the selection goes outside the call for unique() # use things like print or message to check what your function does # print(group_j) # message('this is a message ', group_j) # notice the different color in the console? my_data_i <- example[group == group_j, ] # this would work method_i = method[i] name_1_i = paste0("col1_col_2", method_i) name_2_i = paste0("col3_col_4", method_i) p1_i = stringdistmatrix(my_data_i$col1, my_data_i$col2, method = method_i, useNames = "string") %>% as_tibble(rownames = "a") %>% pivot_longer(-1, names_to = "b", values_to = name_1_i) p2_i = stringdistmatrix(my_data_i$col3, my_data_i$col4, method = method_i, useNames = "string") %>% as_tibble(rownames = "a") %>% pivot_longer(-1, names_to = "b", values_to = name_2_i) p1_i = p1_i[,3] p2_i = p2_i[,3] final_i = cbind(p1_i, p2_i) # results[[i]] = final_i # you replace this content everytime you change groups # you need to append the values between groups (assuming you want one column per test type) # first append, then combine if(j < 2) { # use < instead of == to avoid null error results[[i]] <- final_i } else { results[[i]] <- rbind(results[[i]], final_i) } } }

My code to accomplish the same task

I've added a few checks and balances to make it more dynamic. You can send any number of columns, methods, or groups to grpComp.

This uses tidyverse, glue, and stringdist.

This first function is called by the other function.

library(tidyverse) library(stringdist) library(glue) strD <- function(c1, c2, mm) { # input column 1; column 2; measurement method res <- stringdistmatrix(c1, c2, method = mm, useNames = 'string') f_res <- matrix(res) # extract values and flatten }

This is the distances by group function.

grpComp <- function(fr, methods, grp) { # data frame of columns to compare, # methods to use, groups (vector same length as df rows) cnames <- names(fr) if(length(cnames) %% 2 != 0) { message('there are an uneven number of columns to compare') break # something's wrong } if(length(grp) != nrow(fr)) { message('there groups vector length must match number of rows in the data') break # something's wrong } # extract distances dists <- map( method, function(j) { str_ds <- map_dfc( # by column sets seq(from = 1, to = length(cnames), by = 2), function(i) { str_gr <- map_dfr( # by group unique(grp), function(k) { as.data.frame(list( # has to be list for `col.names` to work strD(fr[grp == k, cnames[i]], fr[grp == k, cnames[i 1]], j)), optional = F, row.names = NULL, col.names = paste0("c", i, i 1, '_', j)) }) # combine groups by rows str_gr }) # combine methods by columns str_ds }) %>% do.call(cbind, .) ncnames <- names(dists) %>% substr(1, 3) %>% unique() # determine unique col groups for(m in ncnames) { # get averages for each comparison set dists <- mutate(dists, "ave_{m}" := rowMeans(across(contains(m))) %>% scale()) } dists <- select(dists, contains('ave')) }

This is how you would use this code.

test5 <- grpComp(example[, 1:4], methods, example$group)

Even though your function for non-grouped data is working, I thought I would include that code as well.

strComp <- function(fr, methods) { # data frame of columns to compare, methods to use cnames <- names(fr) if(length(cnames) %% 2 != 0) { message('there are an uneven number of columns to compare') break # something's wrong } # extract distances dists <- map( method, function(j) { str_ds <- map_dfc( seq(from = 1, to = length(cnames), by = 2), function(i) { as.data.frame(list( # has to be list for `col.names` to work strD(fr[, cnames[i]], fr[, cnames[i 1]], j)), optional = F, col.names = paste0("c", i, i 1, '_', j)) }) str_ds }) %>% do.call(cbind, .) ncnames <- names(dists) %>% substr(1, 3) %>% unique() # determine unique col groups for(k in ncnames) { # get averages for each comparison set dists <- mutate(dists, "ave_{k}" := rowMeans(across(contains(k))) %>% scale()) } dists <- select(dists, contains('ave')) }

To use this function:

test4 <- strComp(example[, 1:4], methods)