FarmCPU output: NA in the fields of allelic effect, SE, and p-value

[AndModica]

Hi,
I've run eight FarmCPU analyses on as many large SNP datasets (~3mln SNPs, 100 individuals; analyses share many individuals, since the panel is a resampling from a larger population). In analysing the results I found that each output has between 23,500-24,400 SNPs with NA in the allelic effect, SE, and p-value field.
Is there any istance in which this can happen?
Thanks.