Academic papers that is:
When I read an article in a reputable journal, I expect that a reviewer has double-checked the results. Am I naive?
I fear that you are. Except in the simplest papers, it is impossible to describe everything that has happened to the data in the text. Only the data themselves can do that, together with the computer code. Reviewers are usually not expected to reproduce all regressions. They usually trust the authors. In contrast, reviewers of theoretical analyses are expected to confirm all equations and proofs.
The pointer is from Marginal Revolution. The situation has been getting better, but estimation remains a laborious business. We'll get there eventually, but at least as far second tier journals are concerned it's not going be in a hurry.