Improving reproducibility in the social sciences: Challenges