Bjelakovic G, Gluud LL, Nikolova D, Whitfield K, Wetterslev J, Simonetti RG, Bjelakovic M, Gluud C. Vitamin D supplementation for prevention of mortality in adults. Cochrane Database of Systematic Reviews 2014, Issue 1. Art. No.: CD007470. DOI: 10.1002/14651858.CD007470.pub3.
Does anyone know if Cochrane publishes the data they use in their meta analysis? I have a suspicion that meta analysis generally does not make good use of the available data. In their vitamin D analysis, they have shockingly large confidence intervals compared to the amount of data they have. I’d like to check that theory.
Does anyone know if Cochrane publishes the data they use in their meta analysis?
I had a look. It turns out Cochrane does publish all their usable data, and they seem to be ungated! Here’s a link to the data for this meta-analysis. (The link to this data is provided in the gated HTML article, but there doesn’t seem to be a link from an ungated page, so I wonder if these data are supposed to be freely accessible… In any case, all their data are currently ungated and accessible by appending ‘/downloadstats’ to the appropriate URL.)
I find myself irritated that they only include effect sizes and sample sizes rather than the actual observed counts for each group, as that would make a Bayesian analysis much easier.
I haven’t looked in detail at it, but is that because their formats or approaches do not support raw data or because they do support raw counts but simply did not supply them? ie they had the data & discarded it, or they may never have had the observed counts & were going off effect sizes reported in papers; the latter is plausible as I’ve found authors very unwilling to share detailed information beyond what is reported in papers.
The link to this data is provided in the gated HTML article, but there doesn’t seem to be a link from an ungated page, so I wonder if these data are supposed to be freely accessible… In any case, all their data are currently ungated and accessible by appending ‘/downloadstats’ to the appropriate URL.
Hm. I wonder how I would get a full list of URLs. It’d be nice to feed it into my archiver bot.
It would be easy to extract a partial list of URLs from this. Google probably has better coverage with its in url search, but I don’t know how to get lots of data out of it.
huh. I didn’t try that because I knew that site: doesn’t work for all prefixes (eg, it fails if you chop off the last digit). I thought it required termination with a slash, but maybe any punctuation works? I do recommend inurl:abstract.
It turns out Cochrane does provide their data. Very nice of them.
Also, at least in this case my own metanalysis based on their data perfectly replicated their results. The inefficiency I thought was there was not there.
Bjelakovic G, Gluud LL, Nikolova D, Whitfield K, Wetterslev J, Simonetti RG, Bjelakovic M, Gluud C. Vitamin D supplementation for prevention of mortality in adults. Cochrane Database of Systematic Reviews 2014, Issue 1. Art. No.: CD007470. DOI: 10.1002/14651858.CD007470.pub3.
Does anyone know if Cochrane publishes the data they use in their meta analysis? I have a suspicion that meta analysis generally does not make good use of the available data. In their vitamin D analysis, they have shockingly large confidence intervals compared to the amount of data they have. I’d like to check that theory.
Here.
I had a look. It turns out Cochrane does publish all their usable data, and they seem to be ungated! Here’s a link to the data for this meta-analysis. (The link to this data is provided in the gated HTML article, but there doesn’t seem to be a link from an ungated page, so I wonder if these data are supposed to be freely accessible… In any case, all their data are currently ungated and accessible by appending ‘/downloadstats’ to the appropriate URL.)
Here are some details about the file formats: http://tech.cochrane.org/revman/documentation/file-formats
I find myself irritated that they only include effect sizes and sample sizes rather than the actual observed counts for each group, as that would make a Bayesian analysis much easier.
I haven’t looked in detail at it, but is that because their formats or approaches do not support raw data or because they do support raw counts but simply did not supply them? ie they had the data & discarded it, or they may never have had the observed counts & were going off effect sizes reported in papers; the latter is plausible as I’ve found authors very unwilling to share detailed information beyond what is reported in papers.
Turns out they actually, do report it! It was just under an unexpected label “EVENTS_1”. I’m going to do a meta analysis of my own.
Followup:
http://slatestarcodex.com/2014/01/25/beware-mass-produced-medical-recommendations/#comment-35322
http://nbviewer.ipython.org/github/jsalvatier/vitamind/blob/master/Vitamin%20D%20meta%20analysis.ipynb?create=1
Hm. I wonder how I would get a full list of URLs. It’d be nice to feed it into my archiver bot.
It would be easy to extract a partial list of URLs from this. Google probably has better coverage with its in url search, but I don’t know how to get lots of data out of it.
Looks like one would be better off using the
site:parameter thaninurl:, since it’s a prefix; sosite:onlinelibrary.wiley.com/doi/10.1002/14651858huh. I didn’t try that because I knew that
site:doesn’t work for all prefixes (eg, it fails if you chop off the last digit). I thought it required termination with a slash, but maybe any punctuation works? I do recommendinurl:abstract.Awesome! Thank you!
It turns out Cochrane does provide their data. Very nice of them.
Also, at least in this case my own metanalysis based on their data perfectly replicated their results. The inefficiency I thought was there was not there.
Sorry, this was an useless post so now it’s gone
Metamed went out of business recently.