This is par for the course with computational research. I discovered a bug in my code the last week before submitting my PhD dissertation. Luckily, all of my code was organized in a pipeline that could automatically regenerate everything on the fly, but I needed a supercomputer. The queue was too long for Titan, so I set up an impromptu cluster on Azure (it was the only cloud provider with Infiniband at the time) and paid $200 out of pocket to regenerate the correct figures.
I wouldn’t be surprised if a significant portion of published computational research has bugs that totally invalidate the conclusions. I think we need to push hard to require all taxpayer-funded research to make any code that results in a journal article publicly available.
I wouldn’t be surprised if a significant portion of published computational research has bugs that totally invalidate the conclusions. I think we need to push hard to require all taxpayer-funded research to make any code that results in a journal article publicly available.