make interesting fake data

[docsteveharris]

take the existing fake data
arbitrarily add some hours to the discharge for patients over a certain age
and then an indicator variable for d/c or not
save that working code for others

this give us our 'test' data to see if we can recover that distinction via bayes