The marvellous Danielle Navarro has LOTS of useful R learning resources on her YouTube channel. This playlist about how to install R and RStudio is particularly useful; no matter which operating system you are dealing with… Dani has you covered.
-
-
-
12 how to install packages
-
You only need to install a package once on your machine. Once the package is installed, you will need to use library() to load the functions in it every time you want to use it, but the installation is a one time job. So you can either do it in the console, or using the packages tab.
-
-
12.1 option 1
-
Install a package by typing the following command with the name of the package you would like to install in the console.
-
install.packages("packagename")
-
-
-
-
12.2 option 2
-
Alternatively, search for the package you would like to install in the packages tab.
-
-
-
-You can search for packages and install them from CRAN via the packages tab
-
-
-
-
Remember once you have installed a package, you will need to use the library() function to load it before it will work.
-
-
-
-
12.3 useful packages for psychology
-
-
tidyverse this is a cluster of super helpful data wrangling and visualisation tools.
-
here this package helps direct R to the correct place for files based on the current working directory.
-
janitor this package helps us clean up data - especially awkward variable names.
-
qualtRics this package is helpful in reading in data files from Qualtrics… except for .sav SPSS format files! (see next)
-
haven this package is a good one for reading in .sav SPSS format files
-
sjplot this package is helpful for making a ‘codebook’ of your variables and values from imported .sav files
-
surveytoolbox this package is helpful in drawing out the value labels of variables imported in .sav format – note: because surveytoolbox is on github and not CRAN, you’ll want to do the following two steps in the console. Note that we do this in the console since we only need to do it once! If the install asks you about updating packages, go ahead and do it! —(1) install the devtools package: install.packages(“devtools”) —(2) install via github: devtools::install_github(“martinctc/surveytoolbox”)
-
ufs this package (short for user friendly science) is a nice tool for computing the internal reliability of scales – note: one of the commands we will use in ufs requires the psych package to be installed (but doesn’t need to be loaded via library()). Ensure you install that first. Two steps: —-(1) install the `remotes`` package: install.packages(“remotes”) —-(2) install via github_lab: remotes::install_gitlab(‘r-packages/ufs’)
-
apa nice for making statistical output into APA style
-
gt nice for making your tables look pretty
-
apaTables makes nice APA-styled tables of correlation, ANOVA, regression etc. output
-
report is a package to help with results reporting
-
psych is an umbrella package for lots of common psych tasks
-
ez is a great package for stats, including analysis of variance
-
emmeans is helpful for comparing specific means in a factorial design
-
-
-
-
-
13 using inline code
-
-
JR maybe this piece needs to go in a separate chapter about writing with RMarkdown, papaja etc
Use of inline code is really helpful in avoiding transcription errors and saving time when writing up! Here, we use code to pull in some descriptive statistics from the exclusion reason table we made above:
-
-
INSERT INLINE EXAMPLE HERE
-
-
-
-
14 helpful console commands
-
-
names(objectname) - returns a list of variable names for that dataframe, making it less likely you will type things incorrectly
-
getwd() - returns the path to the current working directory. Run this in the console.
-
rm(objectname) - removes the object from your global environment. Can be helpful in cleaning up any ‘test’ objects you make while troubleshooting code.
-
?package - brings up the Help info for that package
-
?function - brings up the Help info for that function
-
-
-
-
15 useful keyboard shortcuts
-
Option-Command-I = inserts a new code chunk Command-Enter = runs the chunk of code that your cursor is in
This resource is created in mind so that the community that uses it, can also contribute to it. We hope this mindset will encourage the resource to grow and stay up-to-date for R learners.
-
Importantly, all skill levels are welcome to contribute, even if you think your skills are not up to scratch - this is what this guide is for!
-
-
8.0.1 Hello Quarto!
-
This book is built by Quarto which is an open source, cross-language publishing system that allows users to build beautiful things from blogs, to websites and books!
-
You can learn more about the capabilities of Quarto in this talk by Mine Çetinkaya-Rundel & Julia Stewart Lowndes at posit::conf(2023)
-
-
8.0.1.1 Install Quarto
-
Let’s first make sure we the latest version of Quarto installed.
-
-
-
-
8.0.2 GitHub Workflow
-
Let’s talk about the Git/GitHub part of the workflow. Git is a version control system. It that tracks changes in our code that generated this book and helps us coordinate our collaborative edits.
-
Git monitors the changes in a project directory a.k.a repository or “repo”. There are two types of repositories:
-
-
Local repository: This is a project directory that exists locally in your computer than git is actively watching over.
-
Remote repository: This is a synced copy of your local repo. The remote repo for this book lives on GitHub. GitHub centralises our collaborative efforts. It provides a platform for the community to make a local copy of the book, create changes and introduce them back to the remote repository. GitHub also acts as the hosting service for the book itself which is also a big bonus for us.
Follow these instructions from “Happy Git with R” to install git to your computer. There are specific instructions for different operating systems.
-
-
-
8.0.2.3 Install a git client
-
Traditionally, git has been used in the shell terminal. This is the ‘back-end’ of the computer and you can interact with your computer using code. This means that you have to type out commands to tell git what to do. This method can become a little overwhelming when you are starting out because you have to learn all the different commands.
-
If you are starting out with git and GitHub, we suggest using a git client like Github Desktop, however it is limited to only to Windows and Mac. If you use Linux, GitKraken is also a great alternative.
-
There are many git clients to choose from. Once you start to get the hang on things, all of these clients are much of the same and its a matter of personal preference.
-
If you are an RStudio user, there is a git tab that is built-in to the interface, its features are more simplified than the previously mentioned clients but does the basic commands that we will talk about next.
-
-
-
-
-
-
-Note
-
-
-
-
For the rest of the chapter, we will focus on using Github Desktop, but the main git commands should translate to whichever client you are using.
-
-
-
-
-
8.0.2.4 Cloning: make a local copy
-
Now that we have all the installation out of the way, we can get started!
-
Lets first get you a local copy of the book onto your computer by “cloning” the repository
Here you will find a drop-down menu showing different methods of cloning the code from the remote repo.
-
We will stick with the default option of using HTTPS a.k.a. cloning by using the URL of the remote repo.
-
-
Click on the copy URL button (The stacked squares icon)
-
Navigate to Github Desktop and click on the first drop down menu from the left. Here you will find the Add button.
-
-
-
-
Click on Add and then Clone Repository
-
Click on the URL option and paste in the URL of this book’s remote repo e.g. https://github.com/unsw-edu-au/r4psych
-
Under Local Path, you can choose where you want to place this cloned repo. Avoid putting your repos in Cloud services such as Dropbox or Google Drive. These can sometimes interfere with git’s ability to track your changes.
-
Finally, click Clone and let git do the work! It will copy everything that is in the remote repo and bring it down locally on your computer!
-
Navigate to the local path where you told git to clone to in Step 7 and check out the directories. You will learn more about these in Book Structure
-
-
-
-
8.0.2.5 Branches
-
We will work in branches so as to not overwrite each other’s work and have a systematic way in introducing new contributions to the book. Coordinating collaboratives changes on branches is what GitHub does best.
-
The main branch will be the current approved version of the book. The main branch is what displays at https://unsw-edu-au.github.io/r4psych/
-
A nice clean workflow with branches is to consider them temporary. You pull the most recent from main, you create a branch locally, you make your edits, you commit regularly, you push regularly to github.com, and then you create a pull request for it to be merged into main, and when it’s approved the branch is deleted on github.com and you also delete it locally. That’s the workflow we’ll walk through here. A great resource on GitHub setup and collaboration is Happy Git with R, which includes fantastic background philosophy as well as bash commands for setup, workflows, and collaboration.
-
The following assumes you’ve completed the initial setup from the previous chapter.
When we compute means, we need to set the decimals via round().
-
We also need to tell R to calculate the mean, even if some of the contributing data points are missing. This is what na.rm = TRUE does.
-
As noted above, rowwise asks R to do something for each row (which is what we want here – to compute the mean of the contributing items for each participant). Whenever we use rowwise (or group_by), we need to ungroup() at the end to avoid issues down the line.
-
-
-
-
-
5.4 tables??
-
gt
-
-
gt(scale1_by_condition12)
-
-
apaTable
-
-
-
-
-
-
-
-
-
-
-
-
-
\ No newline at end of file
diff --git a/docs/images/add_repo.png b/docs/images/add_repo.png
deleted file mode 100644
index 6a33352..0000000
Binary files a/docs/images/add_repo.png and /dev/null differ
diff --git a/docs/images/cover.png b/docs/images/cover.png
deleted file mode 100644
index 510fa79..0000000
Binary files a/docs/images/cover.png and /dev/null differ
diff --git a/docs/images/install.png b/docs/images/install.png
deleted file mode 100755
index ae77b87..0000000
Binary files a/docs/images/install.png and /dev/null differ
diff --git a/docs/images/project.png b/docs/images/project.png
deleted file mode 100755
index 9d3b52f..0000000
Binary files a/docs/images/project.png and /dev/null differ
diff --git a/docs/images/readme.png b/docs/images/readme.png
deleted file mode 100755
index 2f2a43a..0000000
Binary files a/docs/images/readme.png and /dev/null differ
diff --git a/docs/images/remote_repo.png b/docs/images/remote_repo.png
deleted file mode 100644
index 2c60fa3..0000000
Binary files a/docs/images/remote_repo.png and /dev/null differ
diff --git a/docs/images/renv.png b/docs/images/renv.png
deleted file mode 100644
index 6ab6b4b..0000000
Binary files a/docs/images/renv.png and /dev/null differ
diff --git a/docs/images/structure.png b/docs/images/structure.png
deleted file mode 100755
index 647e81c..0000000
Binary files a/docs/images/structure.png and /dev/null differ
diff --git a/docs/import.html b/docs/import.html
deleted file mode 100644
index 067d68f..0000000
--- a/docs/import.html
+++ /dev/null
@@ -1,1107 +0,0 @@
-
-
-
-
-
-
-
-
-
-R for Psychology - 3 Import
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
3Import
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
4 Packages for this chapter
-
-
library(tidyverse)
-
-
── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
-✔ dplyr 1.1.4 ✔ readr 2.1.5
-✔ forcats 1.0.0 ✔ stringr 1.5.1
-✔ ggplot2 3.5.1 ✔ tibble 3.2.1
-✔ lubridate 1.9.3 ✔ tidyr 1.3.1
-✔ purrr 1.0.2
-── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
-✖ dplyr::filter() masks stats::filter()
-✖ dplyr::lag() masks stats::lag()
-ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
-
-
library(here)
-
-
here() starts at /Users/fontikar/Dropbox/0 - Job/r4psych
-
-
library(janitor)
-
-
-Attaching package: 'janitor'
-
-The following objects are masked from 'package:stats':
-
- chisq.test, fisher.test
-
-
library(haven)
-library(sjPlot)
-
-
Learn more about sjPlot with 'browseVignettes("sjPlot")'.
Remember the file setup described above? This is where that starts to be important. Remember, our working directory (i.e., where R thinks “here” is) was set via the Rproj file – so it is the “Williams Lab Core R” folder. You can check this by typing getwd() or here() in the console.
-
For most of this core script, we’ll be using data from a file called sampledata.sav, which should be in the data subfolder from the zipped file. If not, sort that out now!
-
A .sav file is in SPSS format. When you export from Qualtrics into .sav/SPSS format, it retains helpful information like question wording and response labels. If you export straight to .csv, you lose that info and will find yourself cross-checking back to Qualtrics. So, strong word of advice to always export to .sav.
-
The code below uses the here command to direct R to the data folder from the working directory, and then the .sav file within it.
-
The glimpse command gives a nice overview of the variables, their type, and a preview of the data.
-
-
data <-read_sav(here("data", "sampledata.sav"))
-
-glimpse(data)
These variable names won’t be very nice to work with with awkward and inconsistent capitalisation. Actual Qualtrics exports are even messier!
-
The clean_names function from janitor helps clean them up!
-
data_cleanednames <- at the start of the line saves the change to a new dataframe. Alternately, you could write it back to the same dataframe (e.g., data <- ), but this should be done very intentionally as it makes it harder to backtrack to the source of an error. The general rule is to create a new dataframe each time you implement a big change on the data.
-
The glimpse command here shows you that you effectively cleaned the variable names!
A few things about working with files in SPSS format (.sav) before we continue. The reason why we bother with this is that the SPSS format maximises the information in the file. Unlike exporting to .csv or another spreadsheet format, .sav retains information about question wording (saved as a variable label) and response labelling (saved as a value label).
-
If you look at the variable types at the right of the glimpse output, you’ll see the some of the variables are dbl (numeric) while some are dbl+lbl (numeric with labelled values). If you view the data object (by clicking on it in the Environment or using view(data)) you will see that some of the variables have the question wording below the variable name.
-
Having this information on hand is really helpful when working with your data!
-
The view_df function from the sjPlot package creates a really nicely formatted html file that includes variable names, question wording, response options, and response labelling. This code saves the html file to the output_files folder using the here package (which starts where your Rproj file is). This html file is nice as a reference for your own use or to share with a supervisor or collaborator!
The data_dict function from surveytoolbox makes a dataframe with all the variable and response labels - similar to the html created above, but this can be called upon later in R as it’s now part of the environment.
Let’s say you just want to know the question wording or response labels for a particular variable, you can do this via code rather than checking the whole dataset. The extract_vallab command from surveytoolbox returns the value labels for a given variable.
# A tibble: 2 × 2
- id demographicscateg
- <dbl> <chr>
-1 1 category 1
-2 2 category 2
-
-
-
There are (evidently) times when packages do not like labelled data. So, here are a few tools for removing labels from the haven package. Keep these up your sleeve for problem solving later! zap_labels and zap_label not surprisingly removes the labels - the first removes the value labels and the second removes the variable labels! The code below makes a new data dictionary of the zapped dataframe and glimpses the new dataframe to confirm the labels are gone.
This book is written in mind for someone working in Psychology and is venturing into R with little to no experience. Research data in Psychology is unique in that it is collected in formats that are human-readable but not exactly R-readable. Data wrangling conventions will often vary depending on research question and therefore by what type data is collected. In “R for Psychology”, we have compiled a series of real-world examples from different sub-disciplines and will walk through the process of wrangling, summary, analyses and visualisations
This book is built using Quarto and hosted via GitHub Pages. It is a living resource, we welcome any contributions from the Psychology community that would improve the quality of this book. There are many ways to contribute:
This book was created on the unceded territory of the Bedegal people who are the Traditional custodians of the lands where the Kensington campus is located.
thoughts from Jenny : I wonder whether it would be useful to structure this in terms of lets combine skills from previous chapters to..
-
-
-
ask a question,
-
get plots/descriptives
-
then answer it with inferentials?
-
And repeat that process for t-test, linear regression and anova?
-
-
-
8.1 Plots and descriptives
-
Now, let’s test whether this scale varies by condition (condition 1 vs. condition 2 in condition12).
-
Before jumping into t-tests, let’s first visualise the data (scale1_index) to get a sense of what it looks like.
-
The first command here tells R to treat condition12 as a categorical variable that groups the data (technically, a factor). This is helpful for plotting. The code uses the dataframe$variablename structure to do a command on just one variable.
-
The second command pipes the data into a command to create a boxplot (geom_boxplot) with the datapoints plotted as well (geom_jitter).The width command makes the plotting of the dots a bit narrower to help with interpretation.
Based on the boxplot, we’d expect a t-test to be nonsignificant - the wide boxes of the plots for each condition overlap, even though the mean of condition 2 is slightly higher than condition 1 (the horizontal line through the box).
-
-
-
8.2 t-test
-
The first line of code uses the base R command to run a t-test. It tells R to do the test on the data_scalescomputed data frame, asking whether scale1_index varies by (~) condition. var. sets the assumption of variance and conf.level sets the desired alpha of the test. That command is wrapped in the handy t_apa function, which makes the output much easier to read (and pull directly into a write-up!), and also allows us to get a confidence interval on the effect size (es_ci).
-
For reporting, you’d probably also want to know the means and standard deviations by condition. The next bit of code asks R to do this, using the summarise function, separately for each condition. Reminder to use na.rm here to account for any missing values.
-
The gt() command shows the requested dataframe in a nice formatting.
## first tell R that condition and demographicscateg are categorical (factors)
-
-data_scalescomputed$condition1234 <-as.factor(data_scalescomputed$condition1234)
-data_scalescomputed$demographicscateg <-as.factor(data_scalescomputed$demographicscateg)
-
-## this creates a dataframe that contains the summary statistics (means, standard deviations, etc.)
-
-scale1_stats <-ezStats(data = data_scalescomputed,
-dv = scale1_index,
-wid = participantid,
-between =c("condition1234","demographicscateg"))
-
-print(scale1_stats)
-
-## this carries out an anova with condition1234 and demographicscateg as between-participants factors
-## the output includes statistical tests of the main effect of condition1234, the main effect of demographicscateg, and the interaction between the two
-
-scale1_anova <-ezANOVA(data = data_scalescomputed,
-dv = scale1_index,
-wid = participantid,
-between =c("condition1234","demographicscateg"),
-return_aov =TRUE)
-
-print(scale1_anova)
-
-## NOTE THAT THE FOLLOWING WOULD BE DONE TO FURTHER INVESTIGATE AN INTERACTION OR A DIFFERENCE IN A MULTI-CATEGORICAL VARIABLE.
-## to build the contrasts we want to do (e.g., baseline vs. midpoint for PhD), we need to ask emmeans to create output to see what the rows represent.
-
-scale1_emm <-emmeans(scale1_anova$aov, ~ condition1234 * demographicscateg)
-
-scale1_emm
-
-## now we can create a set of named vectors that represent each of the 8 means, based on the position in the 8-item list from emmeans.
-
-cond1categ1 =c(1,0,0,0,0,0,0,0)
-cond2categ1 =c(0,1,0,0,0,0,0,0)
-cond3categ1 =c(0,0,1,0,0,0,0,0)
-cond4categ1 =c(0,0,0,1,0,0,0,0)
-cond1categ2 =c(0,0,0,0,1,0,0,0)
-cond2categ2 =c(0,0,0,0,0,1,0,0)
-cond3categ2 =c(0,0,0,0,0,0,1,0)
-cond4categ2 =c(0,0,0,0,0,0,0,1)
-
-cond1 =c(1,0,0,0,1,0,0,0)
-cond2 =c(0,1,0,0,0,1,0,0)
-cond3 =c(0,0,1,0,0,0,1,0)
-cond4 =c(0,0,0,1,0,0,0,1)
-
-
-## now we can ask for contrasts based on these vectors; we will explore the main effect of condition1234 by comparing each of the conditions to one another
-
-scale1_contrasts <-contrast(scale1_emm, method =list("cond1 - cond2"= cond1 - cond2,
-"cond1 - cond3"= cond1 - cond3,
-"cond1 - cond4"= cond1 - cond4,
-"cond2 - cond3"= cond2 - cond3,
-"cond2 - cond4"= cond2 - cond4,
-"cond3 - cond4"= cond3 - cond4))
-
-
-scale1_contrasts
-
-## here's a version with confidence intervals instead of t and p values
-
-scale1_contrastsCI <- scale1_contrasts %>%
-confint()
-
-scale1_contrastsCI
-
-#you may have a specific planned contrast to run. Here let's imagine you want to compare condition 2 to all of the other 3 conditions in one comparison.
-
-scale1_plannedcontraststats <-ezStats(data = data_scalescomputed,
-dv = scale1_index,
-wid = participantid,
-between =c("condition1234"))
-
-scale1_plannedcontraststats
-
-cond2v134 =c(0,1,0,0,0,1,0,0)
-cond134v2 =c(1,0,1,1,1,0,1,1)
-
-scale1_plannedcontrasts <-contrast(scale1_emm, method =list("condition 1 vs all the rest"= cond2v134 - cond134v2))
-
-scale1_plannedcontrasts
-
-
Let’s make a figure!
-
-
data_scalescomputed %>%
-ggplot(aes(x = condition1234, y = scale1_index, fill = demographicscateg)) +
-geom_boxplot() +
-theme_light()
-
-
-data_scalescomputed %>%
-ggplot(aes(x = condition1234, y = scale1_index)) +
-geom_boxplot() +
-theme_light()
This book is organised by data type (e.g. Survey data, Questionnaire data).
-
Each chapter will walk through the process will work with real-world Psychology data and will walk you through reading in data, to cleaning and eventually analysis and visualising the results.
-
-
1.2.1 Conventions
-
We will refer to packages as {dplyr} and functions as mean(). Variables and objects (such as file names or data objects) as age and mtcars. Where it would aid understanding, we will sometimes refer to functions within a particular packages as dplyr::mutute()
-
-
-
-
1.3 Prerequisites
-
Content drawn from existing resources such as https://r4ds.hadley.nz/intro#prerequisites
-
-
1.3.1 R
-
Download Point to intro to R content (RUWithme, Environmental Computing, Software Carpentry)
-
-
-
1.3.2 RStudio
-
RStudio projects Point to resource about Rproj (SWC)
-
Running R code https://r4ds.hadley.nz/intro#running-r-code
-
-
-
1.3.3 Version control with git
-
-
1.3.3.1 What is git?
-
-
-
1.3.3.2 Why do I need git?
-
-
-
-
-
1.3.4 R packages
-
Every code section will always begin with calls to R packages. There will be code that is commented out (have # preceding the code) for you to install these if you don’t have them on your computer
-
-
# install.packages(dplyr)
-
-library(dplyr)
-
-
There are few R packages that will be on heavy rotation when it comes to working with Psychology data. #### {tidyverse}
-
{tidyverse} is a collection of R packages that is essential to a data scientist’s toolkit. By installing {tidyverse} you are actually installing 8 other packages. The ones we will most often use include:
The handy thing is, when you load the {tidyverse} library into R, it will load the core suite of packages for you instead of you loading each of them independently! Efficiency!! 🚀
-
-
library(tidyverse)
-
-
Other packages that will be helpful for your R workflows include:
At the end of each chapter, we will also include our call to sessionInfo() so you can see what version of packages we are using.
-
-
-
-
1.4 Virtual environments
-
Speaking on what package versions as we write this book, we understand the R package space is constantly changing. This means sometimes code will break due to package updates and this is all part of the process! To combat this problem, we’ve enlisted renv to create a reproducible environment for building this book.
-
-
1.4.1 Download our virtual environment
-
The virtual environment used to build this book is stored in a lockfile. You can find this file in the GitHub repository where the source code of this book lives.
-
The lockfile is named renv.lock. You can download this file directly but clicking on the file name and clicking on the “Download raw file” button.
-
-
Alternatively, you can clone our repository into your computer. Learn more about cloning repositorsies and other GitHub workflows in Happy Git by Jenny Bryan.
-
Once you have this file downloaded, move it in a relevant project directory and then we can let {renv} work its magic.
-
-
-
1.4.2 Install
-
First things first, lets install renv if we don’t have it already.
-
-
install.packages("renv")
-
-library(renv)
-
-
-
-
1.4.3 Recreate virtual enviroment
-
Now let’s tell renv where our downloaded renv.lock file is. Specific the path to the file in the function restore() and you are good to go!
There are lots of reasons to use R and R Studio for your data analysis, but reproducibility ranks high on the list. By writing code that reads, wrangles, cleans, visualises, and analyses your data, you are documenting the process that your data has been through from its raw state through to its analysed state. Reproducibility is all about someone else (or you in the future) being able to take your data and the code you have written and use it to produce exactly the same analysis values that you report in your paper or thesis.
-
With this goal in mind, there are a few best practices we recommend, to ensure that your code is usable in the future, irrespective of what machine it is being run on and by whom.
-
-
2.1 R projects
-
R really cares about where things live on your computer, even if you don’t. Humans have gotten out of the habit of thinking very hard about where they put things on their machine; the search capabilities on the modern computer are quite good and you can generally find files quite easily by searching for them.
-
When you are coding in R, however, you need to explicitly tell R where to find things. You can make this process much easier for yourself by always working within an RStudio Project.
-
When you work within an RStudio project, you can reference everything in relation to the top level of that project folder. It doesn’t matter where that project folder lives on your machine (i.e. Desktop, Documents, OneDrive, Dropbox) the code you write is always relative to your project folder. This means that you can share that whole project folder with someone else (your collaborators or supervisor), and your code won’t break.
-
-
When you start a new analysis project, create a New RStudio Project via the File tab.
-
-
-
-
-Working in an RStudio Project called my_new_analysis
-
-
-
Always open RStudio by double clicking on the .RProj file in the folder on your machine. There is an icon in the top right corner of RStudio that shows you which project you are working in and makes it easy to switch between projects.
-
-
-
2.2 where is here?
-
Once you have set up a project to contain your analysis, you can avoid further file path drama by using the here() package. This package makes it super easy to refer to file paths and ensures that your code will work, even if someone else is trying to run it on a different computer.
-
Once you have installed the here() package, use it to tell R where you find your data like this. This code chunk loads there here() and readr() packages and then tells R to read in a csv file, located in the “data” folder, and the file is called file.csv
-
-
mydata <-read_csv(here("data", "file.csv"))
-
-
By referring to the location of your data using the here() package, here is no need worry about working directories, and you can be sure that your code will work on any machine.
-
-
To read more about why projects and the here() package are useful, check out this blog by Jenny Bryan
-
-
-
-
2.3 folder structure
-
Once you have your project set up, you might like to think about imposing some structure on it. It is mostly personal preference, but many analysis projects include the following folders.
-
-
data
-
-
raw-data
-
clean-data
-
-
output
-
-
figures
-
tables
-
-
manuscript
-
-
You always want to keep your raw data untouched and separate from any data that you write back to your machine after your data cleaning process, so a raw-data subfolder can be useful.
-
In addition, you might want to organise your figures and tables into an output folder and put any writing that you are doing in the manuscripts folder.
-
-
If you want to write a manuscript (or maybe your whole thesis!) using RMarkdown, check out the papaja (Preparing APA Journal Articles) package
-
-
Usually the scripts (or R Markdown) documents that you write your code in, live in the top level of your project file. In RStudio, your project structure might look something like this.
-
-
-
-A project structure template
-
-
-
-
-
2.4 naming things
-
When naming things in your analysis folder, it is a good idea to think about your future self. In all likelihood, when you come back to this analysis in a few months time, you will have no recollection how it actually worked, but you can leave yourself some breadcrumbs by being a bit intentional about naming things.
-
Your file names should be meaningful and orderly. The name of each file should tell the new user (or future you) what is in the file and where it goes in the process.
-
-
cleaning-functions.R
-
1_wrangle.Rmd
-
2_visualise.Rmd
-
3_analyse.Rmd
-
-
In this project, I can tell by glancing at the file names that I have a script (.R) that contains functions and three R Markdown files that contain each stage of the analysis.
-
Sticking with lower case is a good idea; avoid special characters and use - or _ to fill any gaps.
In addition to the breadcrumbs that our file names leave, it is also a good idea to leave explicit notes to your future self (or someone else) in a README.md file. This is a simple text file that contains instructions for how the user should engage with your project.
-
Create a new Markdown file and save it as README.md
-
Use it to leave yourself instructions that look a bit like this.
-
-
-
-An example README file
-
-
-
-
-
2.5.2 R Markdown
-
In addition to leaving your future self explicit notes about how to engage with the project generally in a README document, it is also best practice to document your code in a way that makes it really clear what the code is doing and why. For this reason, we recommend using R Markdown documents (rather than R scripts) for your analysis.
-
R Markdown is a handy file format that allows you to intersperse chunks of code with notes. This kind of document makes it easy to write explanations, interpretations, and thoughts for your future self as you code. This kind of documentation makes it much more likely to be able to make sense of what you have done and why, when you come back to your analysis in a few months time.
-
R Markdown documents can also be “knitted” into html, pdf, or word documents, allowing you to share your analysis (and associated thoughts) with collaborators, even if they don’t know R.
