>>12689146>Raw data is available whenever it's reasonable.One of the biggest issues with the replication crisis is that researchers have been unable to provide data, whether it be because they lost it or didn't keep it or something else. And when data does get provided, there are times when it's obvious that the data was handled very poorly or incorrectly. When things like this are kept opaque, the reader assumes by default that things were handled properly. I remember a case where data was kept in an excel spreadsheet, but due to an excel issue, the processing was flawed and the results were spurious. I think some governments ended up partly using that incorrect result to make a fairly large-impact economic decision.
>Ideally there's no point in keeping the raw data since you can always reproduce and confirm the resultsSure, independent replications are better, but that doesn't preclude keeping data. If you are going to have independent replications anyway, then it is strictly better to have kept the data than to not have kept it.
>it's infeasible to keep it around longNot sure how much total data is produced, but many large companies produce and store insane amounts of data these days. And many organizations like labs and observatories already do keep a lot of data on hand. Given that scientific knowledge is very important imo (more important than all the large amount of data that the NSA is collecting for example), I think there's a good case for us to store scientific data collectively.
Further If we used likelihood ratios instead of p-values, it would be much much simpler to combine the data to build up stronger inferences.
