Answer by Alexander Heath for R loop - read and aggregate from matrix - output to new matrix by two unique variables

To do this, I made a function distanceWalked, which calculates the distance traveled for each row except the first.

distanceWalked <- function(data) {  data$distance[1] <- 0  if (nrow(data) > 1) {    for (i in 2:nrow(data)) {      data$distance[i] <- dmatrix[data$section[i-1],data$section[i]]    }  }  return(data)}

I then made a new data table which has all unique combinations of day and id

unique_combos <- unique(data.table(date = dt$day, id = dt$id))

Then I ran a for loop which subsets the data, chronologically orders it by the st column, run distanceWalked on it, and then aggregate it to a new data table

new_data <- data.table()for (i in 1:nrow(unique_combos)) {  dt_sub <- dt[dt$day == unique_combos$date[i] & dt$id == unique_combos$id[i]]  setorder(dt_sub, st)  dt_sub <- distanceWalked(dt_sub)  new_data <- rbind(new_data, dt_sub)}

I then used the dplyr package to find the sum of distance by each unique combination of day and id

library(dplyr)final_data <- new_data %>% group_by(day, id) %>% summarize(total_distance = sum(distance))

It should yield something like this

    day  id total_distance1 02/28 104              32 05/14 104              03 02/26 104              0

This might take awhile to complete for 5 million rows, but it should get you where you need to go!