ds.mean(), ds.var() fail on nfilter for a variable that is NA

datashield / dsBaseClient

DataSHIELD client side base functions

https://www.datashield.org/

GNU General Public License v3.0

11 stars 24 forks source link

Open tombisho opened 3 years ago

tombisho commented 3 years ago

If a variable is completely NA then ds.mean() and ds.var() return the error:

Nvalid less than nfilter.tab

and the call fails. I am not sure that is the right behaviour because that is confusing for the analyst, perhaps they should return NA and not fail?

StuartWheater commented 3 years ago

So the suggested behavior would be:

Column all NA returns: indication "all NA"
Column non-NA above 1 but below nfilter.tab returns: indicates disclosure issue.
Column non-NA above nfilter.tab: return results.

This seem sensible, but I will seek advice.

Stuart

tombisho commented 3 years ago

Yes I think your answer is more complete than mine 😀