Under 10s in New Zealand are being disproportionately hospitalized with Covid
https://fyi.org.nz/request/21573/response/83227/attach/4/HNZ00011574%20OIA%20final%20response%20Hospitlizations%20from%20COVID%2019.pdf
[Perturbation is the adding of random noise to data and suppression is not providing data,
where doing so could compromise confidentiality.
For data that is not already provided exactly on the website:
• All values between 1 and 5 are suppressed
• Zero values are provided exactly, except for in sensitive categories where a zero
value may be suppressed to protect other low values
• Values of 6 and above are perturbed using fixed random rounding to base 3
(FRR3).
This entails rounding values to neighbouring multiples of 3.
For example, 11 may
be rounded to 12 or 9.
FRR3 is a best-practice confidentiality method recommended by Stats NZ.]
--------------------------------------------------------------------------
This is as disconcerting as the data itself.
Stats NZ have a policy reference for the claim of (privacy and best practice confidentiality) on data which is already de-identified when published?
When metrics are small numbers and are suppressed rounded up or down Stats NZ MOH Medsafe know that analysis and calculations become less precise.
Which it can easily be argued is a key purpose of using 'suppression' (s)
https://fyi.org.nz/request/21573/response/83227/attach/4/HNZ00011574%20OIA%20final%20response%20Hospitlizations%20from%20COVID%2019.pdf
[Perturbation is the adding of random noise to data and suppression is not providing data,
where doing so could compromise confidentiality.
For data that is not already provided exactly on the website:
• All values between 1 and 5 are suppressed
• Zero values are provided exactly, except for in sensitive categories where a zero
value may be suppressed to protect other low values
• Values of 6 and above are perturbed using fixed random rounding to base 3
(FRR3).
This entails rounding values to neighbouring multiples of 3.
For example, 11 may
be rounded to 12 or 9.
FRR3 is a best-practice confidentiality method recommended by Stats NZ.]
--------------------------------------------------------------------------
This is as disconcerting as the data itself.
Stats NZ have a policy reference for the claim of (privacy and best practice confidentiality) on data which is already de-identified when published?
When metrics are small numbers and are suppressed rounded up or down Stats NZ MOH Medsafe know that analysis and calculations become less precise.
Which it can easily be argued is a key purpose of using 'suppression' (s)