Scientists aren't always right. In fact, individual research papers turn out to be wrong pretty often and scientists are the first people to tell you that they don't know everything there is to know. They're just working on it with more rigor than most of us.
But scientists are also people. And sometimes, they lie. At Ars Technica, John Timmer looks at some of the most famous cases of scientific fraud and comes away with 8 key lessons that show us how science's biggest scam artists got away with faking their data—sometimes for years.
1) Fake data nobody ever expects to see. If you're going to make things up, you won't have any original data to produce when someone asks to see it. The simplest way to avoid this awkward situation is to make sure that nobody ever asks. You can do this in several ways, but the easiest is to work only with humans. Most institutions require a long and painful approval process before anyone gets to work directly with human subjects. To protect patient privacy, any records are usually completely anonymized, so no one can ever trace them back to individual patients. Adding to the potential for confusion, many medical studies are done double-blind and use patient populations spread across multiple research centers. All of these factors make it quite difficult for anyone to keep track of the original data, and they mean that most people will be satisfied with using a heavily processed version of your results.
3) Tell people what they already know. Since you don't want anyone excited about your work, due to the likelihood they will ask annoying questions, you need to avoid this reaction at all costs. Under no circumstances should your work cause anyone to raise an intrigued eyebrow. The easiest way to do this is to play to people's expectations, feeding them data that they respond to with phrases like "That's about what I was expecting." Take an uncontroversial model and support it. Find data that's consistent with what we knew decades ago. Whatever you do, don't rock the boat.