I always wondered when reading papers how anybody could reproduce the conclusions (p-values, that sort) from the data.

One simple solution would be every time you have some data analysis, you set up a VM or Docker type thing with your dataset and your code. That way other people can download it and run it, see your results, and tinker.