2 ms·
I once heard a great talk by Eytan Domany, a physicist who moved to biology. He had a look at Genome Wide Association studies, where you sequence a number of pe
by adrianN 11y ago
I once heard a great talk by Eytan Domany, a physicist who moved to biology. He had a look at Genome Wide Association studies, where you sequence a number of people with an illness and without the illness, and try to find genes that correlate with having the illness. The problem with these studies, he told us, was that quite often the number of genes they find correlating with having the illness is about the same order of magnitude as the number of participants in the study. But because many biologists are not so proficient at statistics, they don't see that this smells of overfitting their model to the data. He took a particularly popular such study and applied a standard technique to detect overfitting, cross validation, and indeed found that the correlation they reported was spurious.
If I remember correctly, even after he destroyed the study, people kept citing it...