Species Analysis with dplyr and purrr

We are going to use the species data from data carpentry to explore with the dplyr, tidyr and purrr packages. The data are available at http://kbroman.org/datacarp/portal_data_joined.csv. Here we attach the packages we will use with the library command.

library(dplyr)
library(tidyr)
library(purrr)
library(readr)

surveys <- read_csv("../data/surveys.csv")

## Error:
## ! '../data/surveys.csv' does not exist in current working directory:
##   '/Users/brianyandell/Documents/GitHub/Documentation/R/curate'.

surveys %>%
  filter(!(is.na(hindfoot_length) & is.na(weight)),
         !is.na(sex)) %>%
  group_by(species_id, sex) %>%
  summarize(hflen = mean(hindfoot_length, na.rm = TRUE),
            wt = mean(weight, na.rm = TRUE))

## Error:
## ! object 'surveys' not found

Here is a spread table for one of the traits.

surveys %>%
  filter(!(is.na(hindfoot_length) & is.na(weight)),
         !is.na(sex)) %>%
  group_by(species_id, sex) %>%
  summarize(hflen = mean(hindfoot_length, na.rm = TRUE)) %>%
  spread(sex, hflen)

## Error:
## ! object 'surveys' not found

How would you do this for two traits? Turns out this is more challenging, and depends on what you want to compare. The simplest thing is to arrange or subset the earlier table by sex. If you want four columns on one line, for sex and the two traits, you will have to be creative. What you want to do is avoid typing the trait names or the sexes if you can. I don’t have an easy solution!

The following gives us a list of 2-trait tables by sex. Notice the use of map from package purrr to act separately on each sex.

tmp <- surveys %>%
  filter(!(is.na(hindfoot_length) & is.na(weight)),
         !is.na(sex))

## Error:
## ! object 'surveys' not found

tmp <- split(tmp, tmp$sex)

## Error:
## ! object 'tmp' not found

tmp2 <- map(tmp, function(x) {
  x %>% 
    group_by(species_id, sex) %>%
    summarize(hflen = mean(hindfoot_length, na.rm = TRUE),
            wt = mean(weight, na.rm = TRUE))
})

## Error:
## ! object 'tmp' not found

tmp2

## Error:
## ! object 'tmp2' not found

What about this? It gives small tables by species. The challenge is that species that do not have records for both sexes will mess up any easy combining.

tmp <- surveys %>%
  filter(!(is.na(hindfoot_length) & is.na(weight)),
         !is.na(sex))

## Error:
## ! object 'surveys' not found

tmp <- split(tmp, tmp$species_id)

## Error:
## ! object 'tmp' not found

tmp2 <- map(tmp, function(x) {
  x %>% 
    select(-species_id) %>%
    group_by(sex) %>%
    summarize(hflen = mean(hindfoot_length, na.rm = TRUE),
            wt = mean(weight, na.rm = TRUE))
})

## Error:
## ! object 'tmp' not found

tmp2

## Error:
## ! object 'tmp2' not found

Bringing in species information

Suppose you want to bring in taxonomic information from another file. While you can use left_join(surveys, species) to create a larger table, what can we do keeping the two tables separate? This can be useful if both tables are large, or if they are configured in different ways. Here

species <- read_csv("../data/species.csv")

## Error:
## ! '../data/species.csv' does not exist in current working directory:
##   '/Users/brianyandell/Documents/GitHub/Documentation/R/curate'.

Here we want summaries by genus. Notice the use of map_df from purrr to get results as a data frame.

species_taxa <- split(species, species$genus)

## Error:
## ! object 'species' not found

map_df(species_taxa,
       function(x, surveys) {
         # Identify species in this taxa
         taxa_sp <- x$species_id
         surveys %>%
           filter(!(is.na(hindfoot_length) & is.na(weight)),
                  !is.na(sex),
           species_id %in% taxa_sp) %>%
           group_by(sex) %>%
           summarize(hflen = mean(hindfoot_length, na.rm = TRUE),
                     wt = mean(weight, na.rm = TRUE))
         },
       surveys,
       .id = "genus")

## Error:
## ! object 'species_taxa' not found