Why most published research findings are false

So this was an interesting article published in PLoS Medicine by John P. A. Ioannidis. The article lists several reasons of why research is probably false. He argued that the smaller the sample size of your study, the less likely it is that your findings were true. Also, the number of previous studies on related work, the flexibility in the methods, the type of analysis, and the undeclared competing interests all have effects on the probability that a finding is true.

Most of what he said made perfect sense. If you have a study with only seven samples, the likelihood of the samples representing the entire population is fairly small. If you do a study with thousands of samples, your findings have more weight in their statistical significance.

There were several parts of the paper that were frustrating. For example, he didn't seem to have that many suggestions about how scientists can fix this problem. He also thinks that we should declare conflicts of interest much more, which everyone is trying to find significant results so that they can publish and earn more money, be more recognized, get that tenure, or just validate that you are spending all of your time on something significant. No one really goes into research thinking that they aren't going to find anything, otherwise, why would you study it?

He thought that flexibility in methodology makes your results invalid, I think flexibility is what makes your experiment real. Statistics works with theoretically, what is the mathematically perfect way to do this. Reality doesn't work so neat. Cell transformations have a tried and true methodology. And yet, why do so many people have difficulty transforming cells? This is due to the randomness of nature, and the fact that we are dealing with living organisms who react differently to things.

When I first read the article, I freaked out and thought, "So, my entire career is just worthless because most of my findings are probably false?!?" I feel that if so much was actually false, then it wouldn't be repeatable, we wouldn't be able to build off of it, it would appear to be flawed or fall apart. I think this paper is good to remind us that we should be careful and aim to be unbiased and as accurate as possible.