Sample Cleaning Script

We’ve created two versions of a sample data cleaning script, accessible here:

• data_cleaning_template.rmd includes detailed instructions and annotations to help you adapt the script for your own project.
• data_cleaning_template_no_annotations.rmd contains the same cleaning code without instructions or comments, making it better suited for users who are already comfortable with R and want to work with a less visually cluttered script.

These scripts are intended as examples of the data cleaning steps we commonly use for Qualtrics-exported survey data. If you adapt them for your own analyses, be sure to review the checklist below carefully to identify any additional cleaning steps your dataset may require.

Data cleaning checklist

• You have set.seed( ) at the beginning of your cleaning script.

  Note: This ensures that if your cleaning code contains any processes that require random processes (e.g., generating random integers as anonymous Participant IDs), your script will return the same, reproducible results each time. If you’re unsure if your script uses any random processes, set a seed anyways, since it’s otherwise harmless.

• Preview/test responses have been removed

• Participants who were ineligible (because they did not consent or were not old enough to participate) have been removed

  Note: Normally, we program our surveys to direct participants to the end if they are not eligible, and Prolific does not allow participation if you are not 18, so you won’t really have to do anything for this unless you did not program your survey this way or did not use Prolific.

• Incomplete/partial responses removed (according to what you deemed as “enough” responses in your pre-reg)

  Note: Normally, we program survey responses to be Force Response, but give people the option to indicate “Prefer not to say” for demographic questions. If you’ve done things this way, it’s unlikely you will have any “incomplete” responses.

• Duplicate participants removed (e.g., by Prolific ID — people occasionally complete a study twice)

  Note: Normally, we would select “Prevent Multiple Submissions” on Prolific, which would avoid this issue. However, if you collected data through social media sampling, this is worth checking.

• Data has been anonymised, anonymised raw data has been written into a .csv file

Note: Anonymisation entails 1) replacing your ProlificID column with random ParticipantIDs and 2) ensuring comments in open-text responses do not contain any identifying information

• Qualtrics identifying/metadata columns (e.g., “LocationLatitude”, “LocationLongitude”) have been removed
• Column headers renamed to be interpretable
• ‘Prefer not to say’ responses coded as NAs
• Variables have been appropriately coded — categorical variables as factors, numeric variables as numeric
• Participants who failed attention checks have been removed
• Participants who failed bot checks have been removed

Note: When you implement a Captcha question in Qualtrics, they don’t actually allow the participant to proceed in the survey unless they complete it successfully, so this does not require checking. However, if you’ve implemented other bot checks (e.g., authenticity checks on Prolific, enabling bot checks in Qualtrics), you should remove participants who are shown to be highly likely to be bots.

• Sample size before and after exclusions recorded, along with how many exclusions can be attributed to each step
• Any reverse-coding of items that is required has been done
• Pivot clean data to long format (typically required for most analyses)
• Clean data has been written into a .csv file

Before sharing your cleaning code:

• Ensure your code script itself does not contain identifying information (e.g., filtering out participants by ProlificID)
• You have the following files

  Anonymised, minimally processed raw data file Anonymised, cleaned data (you can choose to share a wide and long version separately, if you want to be extra nice) Cleaning script .rmd