I have been reading the excellent Ben Recht on scientific waste. He writes:

No one likes to admit it, but we need bad science to do good science.

This is beautiful rhetoric, but not all kinds of bad science are created equal. Does Recht mean…

• Science that is bad because of small sample sizes?
• Science that is bad because of unrepresentative sampling?
• Scientific studies that are bad because they use questionable methodologies?
• Scientific studies where the authors made errors in their statistical calculations?
• Scientific studies where the authors made programming errors?
• Science that is bad because the authors hypothesized after results were known?
• Science that is bad because the authors performed unethical experiments?
• Science that is bad because the authors fabricated data?
• Science that is bad because the authors are trying to justify discredited ideas?
• Science that is bad because the authors are trying to justify discredited ideas and are also motivated by resentment?

I think there is a line where bad science starts being so bad it starts pushing out the other stuff and it’s between programming errors and HARKing but I’m not even in academia any longer. Still, figuring out where the line is feels important.