We were hoping to test the viability of a new type of statistical test on type 1 and 2 error rates using a trait which has been well-explored with many GWAS hits in the GWAS catalog (e.g. height or BMI). To do this, we would ideally have genotypes from the same platform or imputed dosages for a very large sample that is well-established. This post was asked before here and elsewhere.
Some ideas included the WTCCC data, (though it was done on different platforms and we'd rather not impute,) GIANT (which comes from several investigators and platforms and is not in one place.) But I don't know if they can release imputed dosages. Has anything changed on this? Are there any open genotype-phenotype GWAS datasets? Or must most requests still go through dbGAP? And what is the largest single-stage, single-platform GWAS or single-source for which imputed data is available?
I think most of the answers given to the question to which you linked will still apply. It's hard to get GWAS data unless you generated it yourself; the application process for existing datasets can be frustrating. One approach would be to approach a group who has generated the data directly and collaborate with them, instead of looking for publicly available data (of which I think there is essentially none) or applying to use data as an outsider.
Sounds like a nice idea. In the case of something like GIANT, I have a feeling getting data access would be painful or impossible. Even our group's own large meta-analysis just used ORs/SEs for each SNP as given by other groups. We need at least genotype or dosage data--preferably from a very large dataset.