{"id":21959,"date":"2025-06-25T15:00:00","date_gmt":"2025-06-25T13:00:00","guid":{"rendered":"https:\/\/faces-online.nl\/?p=21959"},"modified":"2025-06-25T15:02:50","modified_gmt":"2025-06-25T13:02:50","slug":"thinking-about-an-experimental-thesis-read-this-first","status":"publish","type":"post","link":"https:\/\/faces-online.nl\/en\/thinking-about-an-experimental-thesis-read-this-first\/","title":{"rendered":"Thinking About an Experimental Thesis? Read This First."},"content":{"rendered":"\n

By Eefje van der Sanden, co-author of Faces Online<\/em><\/p>\n\n\n\n

The What and the Why: Experimental Research<\/strong><\/p>\n\n\n\n

The key difference between experimental research and other types of research lies\u2014perhaps unsurprisingly\u2014in the data collection process. Experiments require primary<\/em> data: data you collect yourself. In contrast, many theses rely on secondary<\/em> data\u2014datasets that already exist and are repurposed for analysis.<\/p>\n\n\n\n

While most students hunt for existing datasets or cobble together various sources, experimental research involves collecting responses through carefully designed manipulations. As the name suggests, experimental research is built around conducting experiments. Simply put, you manipulate one (or two\u2014or if you’re bold, three or more) variables and observe how people respond.<\/p>\n\n\n\n

The fun part? You can draw causal<\/em> conclusions. That is, \u201cIf I do this, then that happens\u201d\u2014as opposed to the more tentative correlations in secondary research, where the best you can often say is, \u201cIf I pull this string, something might happen, but we can’t be sure.\u201d That\u2019s one of the reasons I like experiments: when done well, they let you observe real, measurable effects.<\/p>\n\n\n\n

Now, let\u2019s talk about the word manipulation<\/em>. The first time I heard it in a research context, I\u2019ll admit it sounded a bit sketchy. But in research, a manipulation simply means intentionally changing something in your setup. For example, let\u2019s say you want to study how people react to different versions of a brand message. You could design several versions of the same message\u2014identical except for the wording. One group sees version A, another sees version B. Then, you measure their response in terms of brand trust or brand credibility. Just like that, you\u2019ve tested the impact of message framing. Pretty neat, right?<\/p>\n\n\n\n

One reason I personally prefer experiments is that the data cleaning and analysis are usually more straightforward than in other types of research. Cleaning typically involves removing incomplete responses or those that show patterns of random answering (like selecting the same option throughout). Analysis often involves testing assumptions\u2014like homoskedasticity\u2014and running relatively simple tests, such as ANOVAs or t-tests. That\u2019s it.<\/p>\n\n\n\n

Tips for a Successful Experimental Thesis<\/strong><\/p>\n\n\n\n

If you\u2019re feeling inspired to try an experimental thesis, here are a few crucial tips based on my experience.<\/p>\n\n\n\n

1. Front-Load the Work<\/strong><\/p>\n\n\n\n

If there\u2019s one takeaway from this article, let it be this: preparation is everything<\/strong>. You\u2019ll need to spend significant time designing your experiment, testing it multiple times, getting feedback, and refining your manipulations. With secondary research, you can usually tweak your code and rerun your analysis. But with experiments, once data is collected, it\u2019s locked in\u2014you can\u2019t go back and fix a bad question or confusing manipulation. So make sure your experiment is fully ready<\/em> before you begin collecting responses.<\/p>\n\n\n\n

2. Pre-Test, Pre-Test, Pre-Test<\/strong><\/p>\n\n\n\n

Once you\u2019ve designed your manipulations and questions, it\u2019s time to validate them. A pre-test<\/em> allows you to check whether your manipulations work as intended and whether your measurement items are valid. For example, if you’re measuring brand trust, you’ll likely use multiple items (questions). A pre-test helps you verify if these items are internally consistent\u2014something you can assess using a reliability analysis. You might wonder whether a pre-test is really<\/em> necessary. Technically, no. But if you’re aiming for a robust, credible experiment that measures what it claims to, then the answer is a resounding yes<\/em>. There\u2019s no shortcut here. This isn\u2019t secondary research\u2014you need to put in the work.<\/p>\n\n\n\n

3. Recruit Like It\u2019s Your Job<\/strong><\/p>\n\n\n\n

Let\u2019s talk about the reality of gathering responses. You might think it\u2019s manageable to get 150 people to fill out your survey\u2014and maybe it is. But what if you need 300? That\u2019s a different story.<\/p>\n\n\n\n

Here\u2019s the hard truth: people are busy. Even friends and acquaintances may not be eager to spend seven minutes helping you graduate.  And definitely not that friend you had a class with three years ago. So what can you do?<\/p>\n\n\n\n