Create a loop through a list and add outputs to a dataframe

Question

I want to loop through a list of country names contained world shape file and create individual shapefiles of each country. Then I want perform a calculation on the raster of each shapefile and coerce the results into a dataframe with the country name as an ID variable.

I have written this successfully for an individual country but am struggling to get it to loop correctly.

liech.map <- world.polys[world.polys$NAME == "Liechtenstein",]
plot(liech.map)

rasters <- stack(raster_1, raster_2)
rasters.values <- extract(rasters, liech.map)
df <- as.data.frame(rasters.values)
var <- as.data.frame(weighted.mean(x=df$raster_1, w=df$raster_2, na.rm=TRUE))

What I want to do is extract the list of country names from the world polygon shapefile, create a separate polygon for that country and loop it over every country. Then output 1 dataframe with the `var' in for each country with a country ID.

EDIT

Here is what I've managed to do so far, what I really want to be able to do is feed the following code a list of ID codes/names to loop over. I could of course copy and paste this manually 200-odd times, but this seems such a poor use of time!!

### leichenstein map
## 69.67 sec elapsed
tic()
LTU.map <- world.polys[world.polys$ISO3 == "LTU",]
rasters.values <- extract(rasters, LTU.map)
df <- as.data.frame(rasters.values)
rugged_LTU <- as.data.frame(weighted.mean(x=df$raster_1, w=df$raster_2, na.rm=TRUE))
var_LTU$iso3 <- "LTU"
rm(LTU.map)
toc()

# define master dataframe once
var_master <- var_LTU

### UK map
## 127.31 sec elapsed
tic()
GBR.map <- world.polys[world.polys$ISO3 == "GBR",]
rasters.values <- extract(rasters, GBR.map)
df <- as.data.frame(rasters.values)
rugged_GBR <- as.data.frame(weighted.mean(x=df$raster_1, w=df$raster_2, na.rm=TRUE))
var_GBR$iso3 <- "GBR"
var_master <- rbind(var_master, var_GBR)
rm(GBR.map)
toc()

Answer 1

So first thing first, we create a list to process. world.polys seems to be a data.frame or similar, we want to turn that into a named list.

polys_by_country <- split(world.polys, word.polys$ISO3)

Next we refactor the code for one country into a function:

extract_raster_value <- function(country_map) {
  # Here imagine country map is your LTU.mnap
  rasters.values <- extract(rasters, country_map)
  df <- as.data.frame(rasters.values)
  # compute weighted mean and implicitly return it (last value of function)
  weighted.mean(x=df$raster_1, w=df$raster_2, na.rm=TRUE)
}

OK so extract_raster_value takes a country map and returns a single number, a weighted mean. Note that there is no need to "clean up" the workspace using rm. All local variables defined in the function are only the function scope and do not pollute the global environment.

You can check it works. I have to assume it does, since you have not provided a reproducible example.

LTU.map <- world.polys[world.polys$ISO3 == "LTU",]
extract_raster_value(LTU.map)

The next step is to apply extract_raster_value to each element of polys_by_country.

You can use the apply or lapply functions from base R, but I prefer to use the map family of functions from the purrr package.

library("purrr")

# Apply process_country to each element of the list and return the list of results
map(polys_by_country, process_country)

This returns a named list where the names are the ISO3 names and the values are your weighted mean.

Instead of a list, you can get the result in a named numeric vector with:

result <- map_dbl(polys_by_country, process_country)

This completely avoids loops (or more precisely, hides the loop).

You can easily turn the result into a data.frame if you want:

result_df <- data.frame(
  country = names(result),
  value = result
)

Of course, there may be much better way to do it depending on what is actually in world.polys... Typically, if it is a data.frame, this will be much faster to run:

library("dplyr")
world.polys %>%
  group_by(ISO3) %>%
  summarise(wm = weighted.mean(raster1, raster2))