18 Open Science
Open science refers to a set of practices that make scientific research more transparent, reproducible, and accessible. Adopting open science practices in your thesis not only strengthens the quality of your work, but also aligns it with the standards increasingly expected in Communication Science. I, myself, strongly advocate for the implementation of open science practices (see this paper) and thus expect you to engage with these practices as well.
18.1 Preregistration
Preregistration means publicly documenting your research design, hypotheses, and analysis plan before you collect data. It:
- Prevents p-hacking and HARKing (Hypothesizing After Results are Known)
- Increases credibility of your confirmatory findings
- Distinguishes what you planned from exploratory analyses you add later
Why does this matter in practice? In a landmark study, 100 published psychology findings were independently replicated: only 36% replicated successfully, and the average effect size was half the magnitude of the originals (Open Science Collaboration, 2015). Many of these studies had been reported selectively — only the analyses that produced significant results made it into the paper. Preregistration closes this loophole: if you commit to your hypotheses and analysis plan before seeing the data, you cannot selectively report what looks good.
18.1.1 Where to preregister
The standard platform is the Open Science Framework (OSF): osf.io
To pre-register:
- Create a free OSF account (log in with your institutional email)
- Create a new project for your thesis
- Under Registrations, choose a registration template (e.g., AsPredicted or OSF Prereg)
- Complete your hypotheses, design, and analysis plan
- Submit — you receive a time-stamped, publicly accessible registration
I strongly encourage you to preregister your MA thesis. I can help you how to do it. If you registered, include the OSF link in your method section.
18.1.2 What a preregistration looks like
A preregistration is more than a brief list of hypotheses — it should contain enough detail that someone else could run the study from it. Below is a condensed example for a survey study on privacy literacy and self-disclosure:
1. Hypotheses and theoretical rationale
H1: Privacy literacy will be negatively associated with self-disclosure depth on social media. Research consistently shows that privacy knowledge shapes how people evaluate disclosure risks (Masur, 2020). We therefore expect that users with higher privacy literacy will share less personal information on social media.
H2: The negative association between privacy literacy and self-disclosure will be stronger for users who also report high privacy concerns (moderation). Privacy concerns may serve as a motivational bridge between knowledge and behavior (Dienlin & Metzger, 2016). Without a concern for privacy, knowledge alone may not translate into reduced disclosure.
2. Study design, sample, and power analysis
Design: Cross-sectional online survey.
Recruitment: Participants will be recruited via Prolific Academic. Inclusion criteria: 18+ years old, active social media user (at least once per week), residing in the Netherlands.
Sample size: A power analysis for a small-to-medium correlation (r = .20, α = .05, power = .80) indicates N = 197. We will recruit N = 220 to account for expected exclusions (attention check failures, incomplete responses). We will exclude participants who fail two or more of three attention checks.
3. Measures
Privacy literacy will be measured with the 8-item Online Privacy Literacy Scale (OPLIS; Masur & Scharkow, 2020; e.g., “Websites are legally required to have a privacy policy”). Response format: true/false/don’t know. Higher scores indicate greater privacy knowledge.
Self-disclosure depth will be measured with four items adapted from Bazarova & Choi (2014; e.g., “When I post on social media, I share personal thoughts and feelings”). Responses on a 5-point scale (1 = strongly disagree; 5 = strongly agree).
Privacy concerns will be measured with the 3-item scale by Malhotra et al. (2004; e.g., “I am concerned about threats to my personal privacy online”). Responses on a 5-point scale.
4. Analysis plan
Confirmatory analyses: H1 will be tested with a simple linear regression with OPLIS score as predictor and mean self-disclosure score as outcome (controlling for age and gender). H2 will be tested by adding the OPLIS × privacy concerns interaction term to the regression model. Significant interactions will be probed at ±1 SD.
Exploratory analyses: We will additionally examine whether the associations differ by platform (Instagram vs. Facebook). These analyses are not pre-specified and will be clearly labeled as exploratory in the thesis.
18.2 Open data
Sharing your data and analysis code allows others to verify, reuse, and build on your work.
In your VU thesis, you share data with me via a secure surfdrive folder (see Ethics & Data Management). After your thesis is graded, you may optionally deposit anonymized data in a public repository such as:
- OSF (osf.io) — general purpose, mostly likely the best choice for your thesis
- DANS / Data Archiving and Networked Services (dans.knaw.nl) — Dutch research data archive
- Zenodo (zenodo.org) — CERN-hosted, general purpose
Before depositing any data publicly, ensure all personal and sensitive information has been removed. When in doubt, consult me and the RERC guidelines.
18.3 Reproducible analysis
Reproducibility means your results can be re-created exactly from your raw data and code. Good practices:
- Keep a clean, annotated syntax file (R or SPSS) from the start (see chapter “R for Data Analysis”)
- Use relative file paths so your code works on other machines
- Document every data transformation and exclusion decision
- Consider using R Markdown or Quarto to combine analysis and write-up
For R users: resources are available on the Canvas page R for Social Scientists.
18.4 Transparency in reporting
- Report all tested hypotheses, including non-significant findings
- Clearly distinguish confirmatory (pre-registered) from exploratory analyses
- If you deviated from your pre-registered plan, explain why
- Report exact p values and effect sizes (not just p < .05)
If you want to go beyond the minimum, specification curve analysis (Simonsohn et al., 2020; see also the specr package for R) is an advanced technique for showing how robust your findings are to different analytical choices — for example, different operationalizations of a construct, different sets of control variables, or different sample exclusion rules. Rather than picking one “best” specification, you run all defensible ones and plot the full distribution of results. This makes your analytical decisions transparent and guards against the impression that your findings depend on one convenient choice. It is not required for an MA thesis, but it is a sign of methodological sophistication and increasingly valued in Communication Science.
18.5 Open access
After passing your thesis, you are required to upload your thesis to the UBVU thesis database with public access rights (see Assessment & Graduation). This makes your work openly available for future researchers and students.