@steamtraen.eu
English/Irish/EU in Spain. PhD in psychology & self-appointed data police cadet. Interested in the lower tail of many distributions. Not yet disabled. ID confirmation: https://x.com/sTeamTraen/status/1858181372126408774
I have proposed that we make a data sharing plan part of IRB approval. Progressive IRBs would make sharing the default, and PIs would have to make a specific case for restricting access to the data, and demonstrate why that is more ethical overall than sharing it.
Yes, but for a trial of this nature, proper de-id is not difficult at all. We're not dealing with a total population of 200 vulnerable people who are at risk of being outed or otherwise victimised if someone reverse-engineers their weight and age.
Well, a lot of people think this was a pretty extreme case. 😉
PLOS One's stated policy is "show your data or lose your paper". The EoC is a cop-out.
I have seen IRBs authorise consent statements that say "Nobody outside of the research team will ever see your data", which in effect means "Nobody will ever know if we have made a mistake in the analysis".
I'm not disputing any of that. But the way in which the data are gatekept from peers, who are just as well qualified as the original collectors, often goes well beyond what is reasonable. It also allows authors to get away with fraud, far too often.
poastdoc, shirley
Academics: "Ha ha, this time we've really got capitalism beat"
And frankly, "human subjects protection" is often greatly over-egged. For a study of diet and health, removing the name and perhaps the IP address if they did a survey online is more than enough.
(If the data are all that confidential, who vetted the people in the lab that collected them?)
True, but the code without the data isn't much use anyway.
It's difficult to motivate yourself to make a really good effort to provide high-quality code and documentation when you know that you just made the whole thing up.
Re your second point, you are of course right, but since in this case we strongly suspect that the entire study was fabricated, the fact that the code didn't run might not be the main obstacle to replication.
The problem is that if I have to fix the syntax errors, I am no longer running the same code that the authors did. So any critique I might make of their numbers can be countered with "You must have broken the software when messing with it".
@johnthejack.bsky.social
If authors refuse to share code when asked, post-publication, will you retract the article? Or will you just chicken out and put an eternal Expression of Concern on it, like your colleagues at PLOS One did when authors refused to share their data as "required"? journals.plos.org/plosone/arti...
If authors refuse to share code when asked, post-publication, will you retract the article? Or will you just chicken out and put an eternal Expression of Concern on it, like your colleagues at PLOS One did when authors refused to share their data as "required"? journals.plos.org/plosone/arti...
Most journals have something similar: a policy that looks super-transparent but gives them a get-out clause. Even those with "no exceptions" policies tend not to do anything if the authors later refuse to actually share when somebody asks.
Have a look at the data sharing policy here www.nature.com/nature-portf...
Basically they have to share their data freely unless they can convince the editor that they weally weally don't want to. And the editor will usually go along with that if they want this sexy paper in their journal.
De-identifying the data would take about 10 minutes. After that we would be able to check that they didn't make any very basic mistakes in their analyses. An airliner has two pilots for much the same reason.
The last lot of code I downloaded from the "replication repository" had actual syntax errors in it (mismatched parentheses) and couldn't actually be run. So "It's good enough" is not necessarily true.
*Laughs in European* as LA gas price hits €1.87 per litre
I suppose if Trump does declare victory and stop bombing Iran, he can claim to have stopped yet another war (his 15th?) and Infantino can award him this year's FIFA Peace Prize as well. I don't think anyone has ever won two before.
Is it tho
I'm afraid I didn't get past BOFA without sniggering
Palma harbour at 3pm today. Sometimes reality looks more like AI than AI.
Please say more?
I guess he must be Grumpy
