3 ms·
Statistical Flaw Punctuates Brain Research in Elite Journals
- yamad 12y agoThe problem mentioned--mistreating "nested data" by assuming independence between observations that may actually be linked--is definitely a real one and it's important to bring up, but I don't think it's as severe as suggested here. How did the authors come up with their numbers on which papers had nested designs? I would argue that nested data is on a continuum and it's a matter of judgment and interpretation what to call a singular "research object". If you are studying a single protein molecule, when can you start assuming independence? Protein molecules in the same subcellular compartment? same cell? same cluster of cells? same organ section? same animal? same litter? same species? same animal tech at the animal supply company? same experimental session? same solution stock? It's not obvious where to draw the line. As long as authors are clear and honest about where they decided to draw that line in their analysis, I can assess whether I want to put stock in their conclusions.