Hi MaxRa
Thanks for your question and input. We don't want to encourage too much of a comparison between posts to avoid giving the impression that some are "better" than others. Therefore, we'd prefer to let the post stand as it is.

Not the data, unfortunately. We can share the code though. Please PM me if you're still interested.

We saw no systematic bias on a relevant order of magnitude. So yeah, we mean a null effect. Note the exception in Fig 3A: n=3 for non-binary respondents, where the sample size is very small.