Use skipnan as you would use skipmissing.
SkipNan can be installed from the Julia package registry via:
using Pkg
Pkg.add("SkipNan")using SkipNan
x = skipnan([1., NaN, 2.])
sum(x) # 3.0
argmax(x) # 3
collect(x) # [1.0, 2.0]
collect(keys(x)) # [1, 3]
x[1] # 1.0
x[3] # 3.0
x[2] # MissingException