Grouping & Summarizing

To obtain a summary of a dataframe is useful when trying to get summary statistics or the like from a dataframe. Let’s use an example dataframe that describes the masses of Bison.

Working With Data 

# load libraries
library(lterdatasampler)
library(tidyverse)

bison <- lterdatasampler::knz_bison
head(bison)
## # A tibble: 6 × 8
##   data_code rec_year rec_month rec_day animal_code animal_sex
##   <chr>        <dbl>     <dbl>   <dbl> <chr>       <chr>     
## 1 CBH01         1994        11       8 813         F         
## 2 CBH01         1994        11       8 834         F         
## 3 CBH01         1994        11       8 B-301       F         
## 4 CBH01         1994        11       8 B-402       F         
## 5 CBH01         1994        11       8 B-403       F         
## 6 CBH01         1994        11       8 B-502       F         
## # … with 2 more variables: animal_weight <dbl>, animal_yob <dbl>

It might be useful to understand how the weights of these animals differs by sex. Let’s first check the `animal_sex` column to see what values it takes and what type of data the column is:

str(bison$animal_sex)
##  chr [1:8325] "F" "F" "F" "F" "F" "F" "F" "F" "F" "F" "F" "F" ...
unique(bison$animal_sex)
## [1] "F" "M"

So there are two values, and the column is a character type.

Grouping With One Variable

Let’s try to get the mean and standard deviation for each sex. We could do this by making two different dataframes and then calculating those values, but it might be easier to make this into one separate object that has all the information we need.

To do this we can use the group_by() function from the dplyr package to help us. We will need to group the dataset by the variable we’re interested in, and then we can use the summarize() function also from dplyr to get our values of interest.

To understand how the summarize function works, we can run ?summarize in the console. We see that we will get a new dataframe, with one row for each combination of grouping variables. Here we are only using one grouping variable, and there are two levels to the variable, so we should only have two rows left.

Summarize will make a new column or set of columns, that we can create with base summary functions like summarize(mean = mean()).

Note that when using summarize(), all columns not involved in the grouping will be removed from the resulting dataframe. So, in our case, we will only have one column (animal_weight) left from our original dataframe.

So to group our dataframe and then use summarize() to get our mean and standard deviation for our measures of animal weight, we can pipe these commands together.

mass_by_sex <- bison %>%
dplyr::group_by(animal_sex) %>%
dplyr::summarize(mean_mass = mean(animal_weight, na.rm = TRUE),
                std_dev = sd(animal_weight, na.rm = TRUE))
mass_by_sex
## # A tibble: 2 × 3
##   animal_sex mean_mass std_dev
##   <chr>          <dbl>   <dbl>
## 1 F               762.    282.
## 2 M               728.    420.

Note here that we’ve used na.rm = TRUE to ensure that when mean() and sd() calculate their values, if there are any NA values in the data, they are ignored.

Plains Bison

There are actually two subspecies of bison found in North America, the larger Wood bison (Bison bison athabascae), and the smaller Plains bison that we usually think of, (Bison bison bison). Males are often heavier than females, as this species engages in male-male competition for mate chioce behaviour. The data on organism size in this dataset comes from a long-term study system of a grassland ecosystem at the Konza Prairie Long-Term Ecological Research centre in northeastern Kansas.

Grouping With Multiple Variables

In theory, we can use summarize() grouped by as many variables as we want. To demonstrate this, we can repeat our measurement above, but now also grouping my the month the weight was measured in. We may assume that the animals’ weights will fluctuate throughout the year, and it could be useful to understand how this differs by sex (if at all).

So we will again use group_by() and summarize() to perform this task, but now grouping by both animal_sex and rec_month:

month_sex_weight <- bison %>%
dplyr::group_by(animal_sex, rec_month) %>%
dplyr::summarize(mean_mass = mean(animal_weight, na.rm = TRUE),
                std_dev = sd(animal_weight, na.rm = TRUE))
## `summarise()` has grouped output by 'animal_sex'. You can
## override using the `.groups` argument.
month_sex_weight
## # A tibble: 4 × 4
## # Groups:   animal_sex [2]
##   animal_sex rec_month mean_mass std_dev
##   <chr>          <dbl>     <dbl>   <dbl>
## 1 F                 10      771.    281.
## 2 F                 11      751.    283.
## 3 M                 10      756.    436.
## 4 M                 11      701.    402.

In our little observational look here, we can see that for both sexes, the average weights dropped between October (month 10) and November (month 11).