population normalized read depth

Question

[sumudu_rangika]

Hi,

I have a multi-sample vcf. I filtered variants based on GATK hard-filtering recommendations. E.g. QD<2.0, RMSMapping Quality etc. In my vcf file DP in INFO field have values varying from up to >500. I filtered my variants again based on DP<10 & DP>500. Max cutoff chosen 5-6 standared deviations from the sample mean coverage.(according to GATK). Is this correct?

Or should I normalize the read depth before filter? Which is read depth divided by its chromosomal median. Could someone please explain how to get a population normalized read depth ?

Thanks Rangi