eyeballing AMDB data with lots of spells

top1000.jl

using AMDB
using DataStructures            
using GZip
using UnicodePlots

# replace this line with the path on your own machine
records = GZip.open(deserialize,
                    expanduser("~/research/AMDB/data/AMDB_subsample.jls.gz"), "r")

long_samples = [data for (id,data) in records if length(data.AMP_spells) > 1000]

length(long_samples)            # 6

# let's look at the first one
s = long_samples[1].AMP_spells
unique(spell.employer for spell in s) # about 9 employers
# let's count them
c = counter(Tuple{Int,AMP.Spell})
for spell in s
    push!(c, (spell.employer, spell.status))
end
c.map                           # lots of minor employment spells