4 ms·

Can we have the data to see if their statistics are good and that they applied the right techniques to clean their data? I don't trust any medical study much an

Can we have the data to see if their statistics are good and that they applied the right techniques to clean their data? I don't trust any medical study much anymore because next week one will come out refuting this claim.

metafunctor 11y ago
I think this is the original study: http://archinte.jamanetwork.com/article.aspx?articleid=2091745 http://archinte.jamanetwork.com/article.aspx?articleid=20917...
oliwarner 11y ago
Ask the author. The paper says they had full access to their data. They might be able to pass an anonymised version of that along. They might even be willing. The process of scientific discussion (that you "don't trust") is just how science works. I suspect what you actually dislike is the way these papers are cross-reported. It's one thing to be featured in a relevant journal, it's another to be syndicated as "THE CURE FOR DEMENTIA" by a much higher-level journal (or heavens forbid, a newspaper... or worse). You can blame the sponsors and drug companies for that. Neither of which really seem to be involved here.
- pc2g4d 11y ago
  > The process of scientific discussion (that you "don't trust") is just how science works. The invisibility of the data upon which papers rely is one of the key weaknesses of science as currently practiced, as it makes verification of analysis impossible. I don't think it has to be "how science works", and in fact I hope that science moves away from such opacity to requiring data to be published alongside papers.
Kenji 11y ago
You remind me of Tim Berners Lee in his TED talk: Raw data now! Raw data now! You're right though. Of what worth is a paper without data, without code, without all the relevant resources. It's nothing but an appeal to authority.
- incongruity 11y ago
  Even if you have all that, you still need to trust that the data was collected as described and without any hidden biasing effects – so, again, you're trusting the authority of the author(s)... It's turtles all the way down, in some ways, but I don't disagree with you – code and data are a critical first step and they enable both the verification of findings and better chances at accurately testing the verifiability of the research. It's much better to go out and get your own data and compare apples to apples with the same analytic process (assuming you agree with the analytic process which you should assess even before collecting data, but I digress...)

metafunctor 11y ago

I think this is the original study: http://archinte.jamanetwork.com/article.aspx?articleid=2091745 http://archinte.jamanetwork.com/article.aspx?articleid=20917...

oliwarner 11y ago

Ask the author. The paper says they had full access to their data. They might be able to pass an anonymised version of that along. They might even be willing. The process of scientific discussion (that you "don't trust") is just how science works. I suspect what you actually dislike is the way these papers are cross-reported. It's one thing to be featured in a relevant journal, it's another to be syndicated as "THE CURE FOR DEMENTIA" by a much higher-level journal (or heavens forbid, a newspaper... or worse). You can blame the sponsors and drug companies for that. Neither of which really seem to be involved here.

pc2g4d 11y ago
> The process of scientific discussion (that you "don't trust") is just how science works. The invisibility of the data upon which papers rely is one of the key weaknesses of science as currently practiced, as it makes verification of analysis impossible. I don't think it has to be "how science works", and in fact I hope that science moves away from such opacity to requiring data to be published alongside papers.

Kenji 11y ago

You remind me of Tim Berners Lee in his TED talk: Raw data now! Raw data now! You're right though. Of what worth is a paper without data, without code, without all the relevant resources. It's nothing but an appeal to authority.

incongruity 11y ago
Even if you have all that, you still need to trust that the data was collected as described and without any hidden biasing effects – so, again, you're trusting the authority of the author(s)... It's turtles all the way down, in some ways, but I don't disagree with you – code and data are a critical first step and they enable both the verification of findings and better chances at accurately testing the verifiability of the research. It's much better to go out and get your own data and compare apples to apples with the same analytic process (assuming you agree with the analytic process which you should assess even before collecting data, but I digress...)