dplyr::summarize() V.S. dplyr::summarise()

Background Introduction

As an avid R user for years, I’ve been using tidyverse ecosystem to do my data science work daily and from time to time would come across something new from the tools I use frequently. Last night, as I was building up one of my shiny apps/dashboards by adding test of association of tweets collected from different sampling methods, I found out one of the differences between dplyr::summarize() and dplyr::summarise().

In order to get correlations and p-values, I used a function rcorr() from library(Hmisc). When I added this function into my shiny app, it did not work as it should and my dashboard gave me some strange error message. To better illustrate my points, using mtcars dataset would shed some light on what I would like to convey.

First off, let’s load the respective packages.

library(dplyr)
library(Hmisc)

The following code to me has nothing wrong, as this is the way I usually code, but it has thrown me a strange error. Truth be told, I was confused for a while and did a number of unit testings and checked Hmisc documentation on how the function rcorr() should work, but nothing really stood out and the same strange error appeared over and over.

mtcars %>% group_by(gear, carb) %>%
  summarize(mean(mpg), mean(disp), mean(hp), mean(drat), mean(wt)) %>%
  as.matrix() %>%
  rcorr()

## Error in mean(disp): object 'disp' not found

After searching on stackoverflow and github, I found out this link, which amazed me in a way that I found out a difference between dplyr::summarize() and dplyr::summarise(). To me, they were just the American and British spelling and the code would be doing the same job no matter which one we use, but after I changed z to s for the snippet code, I ended up having the following correct results. Pretty amazing, right?

mtcars %>% group_by(gear, carb) %>%
  summarise(mean(mpg), mean(disp), mean(hp), mean(drat), mean(wt)) %>%
  as.matrix() %>%
  rcorr()

##             gear  carb mean(mpg) mean(disp) mean(hp) mean(drat) mean(wt)
## gear        1.00  0.52      0.25      -0.33     0.27       0.72    -0.53
## carb        0.52  1.00     -0.57       0.33     0.81      -0.04     0.36
## mean(mpg)   0.25 -0.57      1.00      -0.90    -0.79       0.59    -0.92
## mean(disp) -0.33  0.33     -0.90       1.00     0.75      -0.54     0.88
## mean(hp)    0.27  0.81     -0.79       0.75     1.00      -0.24     0.61
## mean(drat)  0.72 -0.04      0.59      -0.54    -0.24       1.00    -0.69
## mean(wt)   -0.53  0.36     -0.92       0.88     0.61      -0.69     1.00
## 
## n= 11 
## 
## 
## P
##            gear   carb   mean(mpg) mean(disp) mean(hp) mean(drat) mean(wt)
## gear              0.1025 0.4500    0.3262     0.4195   0.0132     0.0938  
## carb       0.1025        0.0681    0.3217     0.0025   0.8957     0.2720  
## mean(mpg)  0.4500 0.0681           0.0002     0.0036   0.0551     0.0000  
## mean(disp) 0.3262 0.3217 0.0002               0.0082   0.0843     0.0004  
## mean(hp)   0.4195 0.0025 0.0036    0.0082              0.4866     0.0462  
## mean(drat) 0.0132 0.8957 0.0551    0.0843     0.4866              0.0183  
## mean(wt)   0.0938 0.2720 0.0000    0.0004     0.0462   0.0183

Conclusion

I guess besides spelling difference, there is some underlying structural variations between summarise() and summarize(). Maybe this is a simple note, but sometimes when using summarize() and some strange errors take place, simply changing it to summarise() can sometimes magically solve the issue. Part of the reason why I like data science and do research work in this field enthusiastically is that I always encounter strange errors, some of which are not solved even after checking all components associated with them, but seeing questions and answers posted by people in this collegial community would always give me the right guidance to solving the problems I encounter at any moment. Through conferring and researching the technical questions/problems, I can consistently sharpen my data science tools.