Disclosure control

[tombisho]

Look at Swiss knife disclosure measures and implement them

[tombisho]

and check how the seed is set

[tombisho]

Note that a variable to be synthesized
first that has no predictors is a special case and its synthetic values are by default generated
by random sampling with replacement from the original data ("sample" method). I

[tombisho]

• concern that the density smoothing does not hide extreme values, so might need to use top and bottom coding. How to set the top and bottom? 90% of real value? Might hide large values causing problems?
• add a label to the data to show it is synthetic (easy?)
• remove unique combinations of factors that are also in the real data
• first column always sampled - contains real data? Force it to be a factor? The smoothing and top and bottom is applied to it, which can hide the real values