In this post, I want to answer the question: Is it necessary to replicate the results of a published paper before publishing another paper on the same topic? To start with an answer, we first need to dive into the question, what is a replication study? It’s not the case that there are either replication studies and non-replication studies. There are many shades in between.
When you talk about replication, you could mean a very direct replication, essentially redoing everything that a study has done. That is not happening very often. But what we see much more often is a partial replication. Let’s say there is a model you want to explain x with factors a, b, and c. A direct replication would be the same model, the same context. But maybe you’re doing something that tests the same model in a different context. That isn’t a direct replication, but it’s quite close. If there’s nothing that speaks against it, it makes your results more credible because they have been partially replicated.
The same goes for model changes. Say you’re now trying to explain x through factors b, c, and d. You’ve left out a and added d. But you have some information about the effects of b and c. If there’s some outcome that aligns with previous studies, that is an outcome that replicates what has been done before.
To come back to your original question, whether you need a replication before publishing another paper on the same topic, no. The way it usually works is that you do not start with something completely different. You go in the direction of a partial replication. You are building on what is already there, using models, methods, contextual information as you have it available, and adjust it slightly. You arrive at something new that can also be easily tied to something that already existed.