3 ms·

Randomized controlled studies have little to no scientific value. Too easy to report random results, impossible to control for all factors, sample sizes never r

Randomized controlled studies have little to no scientific value. Too easy to report random results, impossible to control for all factors, sample sizes never remotely large enough. Psychiatric theories can’t be falsified, making them not science.

ndidoiul 9y ago
Ok... so here's another tidbit. Meta-analysis was really developed (although not originated) in psychology, attempting to wrestle with replicability issues in much the same way as is the case now. It's really the introspective nature of the human behavioral sciences that drives that as much as the complexity of what's being dealt with. And I'm not sure why you think psychiatric theories can't be falsified. It happens all the time. If it couldn't, this registered replication stuff wouldn't be getting so much attention, because it would be a non-starter.
- valuearb 9y ago
  Meta-analysis can't solve the problems with statistical studies. Disparate small studies don't correlate well with a single larger study, given different standards for measurements, participants, different and possibly overlapping pools, etc. Meta-analysis also suffers from publication bias, agenda based bias, etc. How does running a bunch of replications of a study ever falsify it? Based on statistical probability it's always possible the original study was accurate, and it's the replications that cheated. The realization that the results of these studies are essentially random goes a long ways towards establishing how unfalsifiable any of them are. And beyond statistical studies into analytic theories, how do you falsify the theory of the "Id, ego and super-ego"? Transfer neuroses? How do you falsify "Dialectical behavior therapy"? It worked for someone somewhere, so it must be true. The only thing that's come out of the whole of psychiatry and psychology that has any scientific merit is Skinners behavioral experiments.
dm319 9y ago
> Randomized controlled studies have little to no scientific value Really? How would you convince me that, say, Nivolumab is a better treatment for cancer than the current standard of care? I doubt you have something better than a randomised controlled trial, given there is international acknowledgement that this is the gold standard - the only thing to top it is several RCTs spanning several centres and countries. >Too easy to report random results Subgroup analysis is always taken with a pinch of salt. >impossible to control for all factors Well that's the point of a randomised controlled trial. The 'controlled' part is 'randomised'. >sample sizes never remotely large enough For what? >Psychiatric theories can’t be falsified, making them not science. Not sure what question you have in mind, but the question "Does anti-psychotic medication 'A' improve overall survival in patients diagnosed with refractory schizophrenia" is certainly a question that can be tested in a scientific way. I'm not a psychiatrist, but your comments don't make sense.

ndidoiul 9y ago

Ok... so here's another tidbit. Meta-analysis was really developed (although not originated) in psychology, attempting to wrestle with replicability issues in much the same way as is the case now. It's really the introspective nature of the human behavioral sciences that drives that as much as the complexity of what's being dealt with. And I'm not sure why you think psychiatric theories can't be falsified. It happens all the time. If it couldn't, this registered replication stuff wouldn't be getting so much attention, because it would be a non-starter.

valuearb 9y ago
Meta-analysis can't solve the problems with statistical studies. Disparate small studies don't correlate well with a single larger study, given different standards for measurements, participants, different and possibly overlapping pools, etc. Meta-analysis also suffers from publication bias, agenda based bias, etc. How does running a bunch of replications of a study ever falsify it? Based on statistical probability it's always possible the original study was accurate, and it's the replications that cheated. The realization that the results of these studies are essentially random goes a long ways towards establishing how unfalsifiable any of them are. And beyond statistical studies into analytic theories, how do you falsify the theory of the "Id, ego and super-ego"? Transfer neuroses? How do you falsify "Dialectical behavior therapy"? It worked for someone somewhere, so it must be true. The only thing that's come out of the whole of psychiatry and psychology that has any scientific merit is Skinners behavioral experiments.

dm319 9y ago

> Randomized controlled studies have little to no scientific value Really? How would you convince me that, say, Nivolumab is a better treatment for cancer than the current standard of care? I doubt you have something better than a randomised controlled trial, given there is international acknowledgement that this is the gold standard - the only thing to top it is several RCTs spanning several centres and countries. >Too easy to report random results Subgroup analysis is always taken with a pinch of salt. >impossible to control for all factors Well that's the point of a randomised controlled trial. The 'controlled' part is 'randomised'. >sample sizes never remotely large enough For what? >Psychiatric theories can’t be falsified, making them not science. Not sure what question you have in mind, but the question "Does anti-psychotic medication 'A' improve overall survival in patients diagnosed with refractory schizophrenia" is certainly a question that can be tested in a scientific way. I'm not a psychiatrist, but your comments don't make sense.