ABSTRACT
Lack of replicability in computational studies is, at base, a problem of shortcomings in record keeping. In laboratory-based experimental science, the tradition is to write down all experimental details in a paper notebook. This approach is no longer viable for many computational studies as the number of details that could have an impact on the final result is so large. Automated or semiautomated tools for keeping track of all the experimental details—the scientist’s own code, input and output data, supporting software, the computer hardware used, etc.—are therefore needed.
