names were over-written

10%, 50%, 100% (not a sample!!)

explained sub-sample selection process

checked for transitions between different ‘valid observation’ states

added do_graphs switch
checks for valid gas values of 0 due to rounding of 100 < gcons < 250
to ‘nearest’ 500 - i.e 0!

2012 descriptives uses 20pc sample file
Long file descriptives loops over % samples

set the % you want using the ‘sample’  local. By default this
constrains the selection to keep the overall ratios of EE_BAND
PROP_TYPE constant (as in the original sampling)