It seems to me that there are many more meta analysis published now than were done say 20 years ago. I assume that statistics has not evolved. I suspect that the cause might be the proliferation of predatory journals, or perhaps I am just not remembering things accurately. Has anyone done a study of this?
So you are looking for a study of studies of studies, huh?
You need to have studies before you can do studies of studies (and studies about studies of studies and…). There have been lots of studies done in the past 20 years.
It is true that there has been, but honestly there were plenty before that as well. It does not explain why (assuming they truly have increased).
Another thought is that perhaps some statistical tool has become accepted in the interval to explain this increase.
This is a very controversial subject. Properly conducted meta-studies accurately reflect the results of prior studies. Whether or not that is broadly meaningful is hotly debated. So if you are interested in characteristics of the prior studies then a meta-study can be useful. If you are drawing conclusions about the subject of the original studies you end up with a lot of arguments.
ETA: The expansion of this technique in the last 20 years has a lot to do with more readily available information and analytical tools.
I suspect it has to do with the advent of computers and data reporting requirements. Before in order to get data from multiple studies you would have to contact multiple authors and get them to send you their data, which would involve a great deal of effort to them with little payoff. Now days journals generally require that the data be made publicly available online so getting data from multiple studies just involves a few mouse clicks.
That’s a pretty questionable assumption.
Well, as folks like Nate Silver have shown in pooling polls and accurately predicting the last US elections, pooling the results of studies gives you a larger sample. If the original studies don’t have flaws and the aggregator doesn’t cherry pick studies, then aggregating studies should lead to better and more confident results.
It’s also a way to gather disparate studies, some of which may have never gotten the notice they deserved. It also serves to answer the question: Do we need a newer and better study on this issue, or has enough work already been done?