Brain studies show thousands of participants are needed for accurate results

[Genady]

My suspicions have been confirmed. When I studied cognitive neuroscience and when I read news in this area, it was and is very disconcerting to see how often far fetched conclusions about causation are made there on a shaky basis of a small sample size, a multitude of uncontrolled parameters, and correlations only. This new study confirms that more often than not these results are unreliable and unreproducible. A new research paradigm is needed to move this science from stagnation.

Quote

The researchers found that brain-behavior correlations identified using a sample size of 25—the median sample size in published papers—usually failed to replicate in a separate sample. As the sample size grew into the thousands, correlations became more likely to be reproduced. 

Brain studies show thousands of participants are needed for accurate results | University of Minnesota (umn.edu)

[swansont]

I think statisticians have known this for a relatively long time.

[Genady]

Yes, my professor in biological statistics was very explicit about these studies.

Here is a recent example of a study with far fetched, unwarranted, and completely speculative conclusions:

Quote

To study this, Rutishauser and his colleagues worked with 20 patients who were undergoing intracranial recording of brain activity to guide surgery for treatment of their drug-resistant epilepsy.

Researchers uncover how the human brain separates, stores, and retrieves memories | National Institutes of Health (NIH)

[CharonY]

Especially with high-dimensional data (incl. MRIs) associations are often spurious. At the same time, only allowing high-participation studies to exist would effectively eliminate smaller labs and result in loss of creativity as studies would only be conducted by a few (successful) folks.

[Genady]

A "darker" possibility is that MRI scans just don't represent well the underlying factors in the brain that affect behaviors. IOW, MRI scans might be a wrong measure, like scull shapes in phrenology.

[Dropship]

Regarding sample numbers, aluminium has been linked to Alzheimers but I'm not sure if all victims have aluminium flecks in their brains or just a few.

But on the assumption that there's no smoke without fire, I decided to play it safe and switched from aluminium pots and pans long ago-