Masking regression data

[cd-mj1124]

I would like to mask the observations in my regression data set, but would like the coefficients/standard errors to remain constant between regressions run on the masked and unmasked data.  Is this possible?  How would I go about doing this?  I think I would have to do something with the correlations between the variables, but not sure where to go after.  Thanks.

[Reeza]

What do you mean by mask the observations?

[cd-mj1124]

I have data that is confidential.  I would like to visually change the values so that I could publish the data set, but also preserve the regression coefficients/standard errors and underlying correlation structure of the original, confidential data.

[Reeza]

That's not a trivial problem - in fact its incredibly difficult. (I would love to be proven wrong on this topic)

Here's Statistics Canada methodology.
https://www12.statcan.gc.ca/nhs-enm/2011/ref/pumf-fmgd/P3-eng.cfm

[PGStats]

I think you could reach some level of anonymity by removing axis scales on graphs and by removing intercepts in regression equations.