Why does a SNP have to be in a peak to be a legit GWAS result.

Question

[Tom]

I notice that regularly, when authors are describing the relationship between a GWAS peak and a trait. Why does the SNP of interest have to be in a peak? Would a SNP that showed a significant association with a trait, but was not in a peak, be viewed as a spurious result? Thanks

[karl.stamm]

What's a peak? If you can figure out what a running window of LOD is, then there's your answer.