Khan Academy video on Statistical Power

Group summary statistics with the aggregate() function

Getting a summary statistics based on groups (e.g. a categorical variable) is a common activity in data analysis. One way to do this is in R is to make individual subsets of the dataframe for each level of the group using the subsetting methods we have used in previous modules. However, this can become arduous for groups that have many different categories, and can end up filling up your memory if you’re working with big data. A better approach is to use the aggregate() function. Here’s an example using the mtcars dataset, where we calculate the mean hp for each type of cyl:

# Average horsepower for each cylinder type
aggregate(hp ~ cyl, data = mtcars, mean)

##   cyl        hp
## 1   4  82.63636
## 2   6 122.28571
## 3   8 209.21429

We can also save this output into a new object, and subset parts of the new dataframe to make comparisons:

# Average horsepower for each cylinder type
avg_hp_cyl <- aggregate(hp ~ cyl, data = mtcars, mean)

# On average, how much more horsepower do 6 cylinders than 4 cylinders?
avg_hp_cyl[2,2] - avg_hp_cyl[1,2]

## [1] 39.64935

Note that this is not only limited to calculating means, we can use other functions like sum, min, and max.

Group summary statistics with the dplyr package

Another way to do the same calculation is to us the group_by() and summarize() functions in the dplyr package. This approach is nice because we can use the pipe operator %>%, and it also works faster for large datasets than the base R approach:

library(dplyr, quietly = T)

## Warning: package 'dplyr' was built under R version 3.5.1

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

mtcars %>% 
  group_by(cyl) %>% 
  summarize(avg_hp = mean(hp))

## # A tibble: 3 x 2
##     cyl avg_hp
##   <dbl>  <dbl>
## 1     4   82.6
## 2     6  122. 
## 3     8  209.

Other notes

• Information on the final exam schedule will be coming shortly.

• If the Console pane in RStudio showing a + that means that R is expecting more information, which typically means that there is a missing " or ).

• Use the == (which reads as “is equal to”) when subsetting, not the = (which is using for assignment).

• Don’t worry if you’re having trouble defining p-values, many scientists and researchers do too!