How does a researcher ensure that reliability is maintained in instruments, and between raters?