Will you remove some of the obvious fake outliers from the data set? eg. Line 338 on the .csv file is a 99 year old female self-employed, pre-high schooler with 10+ kids working 140 hours per week and earning $1M per year. Sounds a little suspect to me...
Not to mention the 22 year old student who completed the questionaire 4 times and has 70 years of experience, works negative hours and earns $100M.
Or the divorced 407 year old female with 700 years of experience and 8 kids.