Think about issues related to NO epidemiological data

[timo-b-roettger]

Some unordered thoughts:

• I think it would useful to involve at least 1 or 2 experts that have worked with these kind of data from different angles
• the uniqueness of dataset and high impact on research landscape is a good argument to generate broad interest
• as soon as we have the data, we can approach the press to broadcast our message

A possible workflow could look like this:

• identify automatic way to identify target articles (e.g. by search string related to data set in abstract or title using web of science)
• identify target estimands (e.g. open raw data, open processed data, open scripts, open materials/protocols, preregistration)
• if the sample is really large, we can do a two-part analysis. (a) An automated analysis on the whole sample, screening for target search strings and (b) a manual analysis, screening random sample of articles for cues to transparent practices.

[julienmayor]

Good points - it would be great to include other metrics indicative of bias (distribution of p-/z-values, p-curve analyses, etc). That would supplement the documentation of openness (i.e., if it isn't open but nothing is indicative of bias then maybe not a severe issue?) Cf experts - any ideas?

[timo-b-roettger]

Can someone link to the data bases? we should reach out to colleagues and see who can help with this. I would need more infos to reach out. @julienmayor

[julienmayor]

Here is the cohort that I had in mind: https://www.fhi.no/en/studies/moba/ - 950 articles published as of last month...

[timo-b-roettger]

Given the strict publication guidelines, these should be easy to identify really.

[julienmayor]

Absolutly!