The sad truth is that the majority of social sciences likely have similar problems surrounding data integrity. Good data is often expensive or hard to acquire,

The sad truth is that the majority of social sciences likely have similar problems surrounding data integrity. Good data is often expensive or hard to acquire, and once you have it, you often have to apply some sort of ad hoc interpretation in order to make it quantifiable for performing statistics. E.g. in psychology there are entire manuals for how you "encode" interactions between a therapist and a patient along various dimensions. And then once you've got quantitative data, many social scientists don't actually apply statistics correctly, either out of ignorance or willful desperation to have some sort of positive result. We probably need to have to have some sort of standard for what constitutes acceptable quality data, and have a government funded repository of such data to encourage reproducibility. But that may be way easier said than done. I think I'll just stick to mathematics.