Motivation and goals
This semester, we’ve covered a lot of material about how to clean, wrangle, and visualize data in R. These are important foundational skills for whatever coding you might need to do in the future. However, they won’t be enough. Inevitably, you will hit up against challenges in R – the code you wrote might not run, or you might not be able to figure out the right steps to follow to get your data into the format you want for a particular analysis. This is where de-bugging and searching skills come in. One of the most useful things you can learn about coding is how to problem-solve when your code isn’t doing what you want it to. In this week’s session we’ll give you some tips for these situations. We’ll start by talking about how to figure out where you went wrong if your code doesn’t run. Then, we’ll talk about how to search for answers when you aren’t able to figure out a coding problem on your own. Finally, we’ll discuss some common problems to be aware of in R.
Throughout this lesson, we’ll use examples from the fivethirtyeight package, as we did in Office Hours 7.
Let’s start by loading the tidyverse and the fivethirtyeight package:
library(fivethirtyeight)
library(tidyverse)
Figuring out where you went wrong
A first step in de-bugging R is to try to figure out where you went wrong. We’re going to discuss a few different ways of doing this.
Locate the bug within your code by running small sections of code one-at-a-time.
Start from the top of your code, and run small segments one-at-a-time. Alternatively, if you have no idea where the error is located in a long piece of code, run the first half of your code to see if it is there, and then gradually narrow down into the segment where the error is located.
If your code is formatted as a for-loop, set i to a particular value, and then run each line of the for-loop for that value of i. This will allow you to see which line of the loop caused a problem.
Look at the data you’re working with (or, at least some of it)
We’re going to work with the the dataset comic_characters from the fivethirtyeight package as an example.
If you want to look at the data, you have several options:
dat <- comic_characters
# Look at the whole dataset
View(dat)
# Look at a few rows of the dataset
head(dat)
## # A tibble: 6 x 16
## publisher page_id name urlslug id align eye hair sex gsm alive
## <chr> <int> <chr> <chr> <chr> <ord> <chr> <chr> <chr> <chr> <chr>
## 1 Marvel 1678 Spid… "\\/Sp… Secr… Good… Haze… Brow… Male… <NA> Livi…
## 2 Marvel 7139 Capt… "\\/Ca… Publ… Good… Blue… Whit… Male… <NA> Livi…
## 3 Marvel 64786 "Wol… "\\/Wo… Publ… <NA> Blue… Blac… Male… <NA> Livi…
## 4 Marvel 1868 "Iro… "\\/Ir… Publ… Good… Blue… Blac… Male… <NA> Livi…
## 5 Marvel 2460 Thor… "\\/Th… No D… Good… Blue… Blon… Male… <NA> Livi…
## 6 Marvel 2458 Benj… "\\/Be… Publ… Good… Blue… No H… Male… <NA> Livi…
## # ... with 5 more variables: appearances <int>, first_appearance <chr>,
## # month <chr>, year <int>, date <date>
# Look at the variable names
names(dat)
## [1] "publisher" "page_id" "name"
## [4] "urlslug" "id" "align"
## [7] "eye" "hair" "sex"
## [10] "gsm" "alive" "appearances"
## [13] "first_appearance" "month" "year"
## [16] "date"
# Look at a few variables of the dataset
View(data.frame(dat$publisher, dat$year, dat$name))
Clear everything and run your code again
Something might have gone wrong if you ran code out-of-order or you ran the same lines multiple times. In these cases, it can help to clear everything in your environment (by clicking on the broom) and then run your code again.
You can also clear the environment by running rm(list=ls()).
Restart R
You may be inadvertently using some variable that you saved globally. Try re-starting R to see if your code will run without an error.
If you have complex functions, try using options(warn = 1)
Normally, if there is a warning, R will print this when it’s done running the function. If you want to have warnings printed as they occur, you can specify this in a function.
Make a mini example
If you’re not sure how R is getting to the (incorrect) calculations it’s giving you, try running the code on a small subset of your dataset so that you can calculate things by hand and compare to the R results.
Common problems
Something went wrong with missing values.
You might be having issues with missing values in your code. One thing I commonly do to check missings is I count up how many missings are in each row and each column of my dataset. You can do this as follows:
# To count up the missing values in each row:
dat$missings_byrow <- apply(dat, 1, function(x) sum(is.na(x)))
table(dat$missings_byrow)
##
## 0 1 2 3 4 5 6 7 8 9 10 11
## 81 5401 7180 5316 3260 1323 436 157 51 50 14 3
# To count up the missing values in each column:
missings_bycol <- apply(dat, 2, function(x) sum(is.na(x)))
missings_bycol
## publisher page_id name urlslug
## 0 0 0 0
## id align eye hair
## 5783 6186 13395 6538
## sex gsm alive appearances
## 979 23118 6 1451
## first_appearance month year date
## 69 815 884 886
## missings_byrow
## 0
It may also be helpful to look at the rows where a particular value is missing in order to understand what is going on. To do this, you can filter and look at the dataset:
dat_check <- filter(dat, is.na(dat$align))
View(dat_check)
The function you are using is part of multiple packages in R.
In some situations, a particular function name may be used by multiple packages in R. If you have loaded both of these packages, R won’t know which one to use. In these situations, you can type the name of the package you’re interested in using followed by :: (e.g. dplyr::) before the function name to ensure R uses the function you want it to.
Other issues:
- Mismatching parentheses
- Using the wrong number of
= signs
- Giving arguments to functions in the wrong order, and not naming them
Exercises
We’re going to start with the dataset classic_rock_song_list. This dataset contains the data used in this article. We’re going to try to recreate a table from the article, showing the 25 most commonly played songs across 25 classic rock stations:
The correct code to make this table is included below, but don’t look ahead…the point of this lesson is to learn how to troubleshoot, not to look at the answers!
Hint: start by loading the dataset into R:
dat <- classic_rock_song_list
# Load the data
dat <- classic_rock_song_list
# Remove the observations that are missing release years, and organize in order of playcount
fortable <- dat %>% filter(!is.na(release_year)) %>% arrange(desc(playcount_has_year)) %>% select(song, artist, release_year)
# Select the first 15 rows
fortable <- fortable[1:15,]
# Look at the results
View(fortable)
Next, let’s look at the dataset US_births_1994_2003, which was used in this article. We’re going to try to recreate this figure, showing that people are less likely to have babies on Friday the 13th:
Hint: start by loading the dataset into R:
dat <- US_births_1994_2003
I haven’t included the answer here (because I haven’t fully figured it out myself!), but this is a good example to practice on.
