In the Styrkarsdottir et al paper the p-value they used to indicate statistical significance at the genome level was p ≤ 1.7 x 10-7. Why did they set their p-value for significance so low? Why didn’t they use the p ≤ 0.05 value that is commonly used in statistics?
The significance was set so low because the study was testing for an association between 301,019 SNPs and the bone mineral density of the hip and lumbar spine. All these 300,000+ SNPs, in order to be reported as statistically significant, needed to be accounted for in order to get the genomic wide significance. The p-value was then taken as 0.05 (standardly used p-value in stats) DIVIDED by 301,019 SNPs in order to give this genomic wide significance.