When seeking advice on good programming practices the typical answers are a variation of agile software development, test driven development or design patterns. However, as far as I know, neither of these are proven using the scientific method (if I'm wrong on this then feel free to correct me).
I wonder, are there any good resources on the topic of evidence-based development practices?
There has been some research done on the impact of particular agile practices, but it is definitely sketchy. There's likely a similar level of work with design patterns.
The links I'm putting here are by no means definitive, but have some value, and also reference other research that might be worth looking at.
Part of the value in reading these will be an understanding of how complex and difficult a question you have asked.
Really the most important evidence will be supplied by your own experience, should you be convinced to try some of these practices. It will also be difficult for you to quantify, but you might be able to decide whether the practices work for you.
If you are really interested in this then the best source of information would be academic journals or the IEEE as donroby mentions. If you have a degree from a uni and are an Alumni member still then you may be able access the information through the university library (At least that is how it works in Australia. Alumni can get cheap membership to the library at the uni and access to a lot of the resources students use.)
EDIT: Most of the information will not be experimental but studies of projects and surveys. You can't really apply the scientific method in this kind of thing; it's closer to social science than it is the hard sciences like chemistry and physics