I have a dataset of N
items, with attribute V
which can be repetitive, for example:
City | Country
--------------
NYC | USA
LA | USA
LON | UK
MAN | UK
AMS | NL
After bootstrapping the dataset into M
groups on size N
with replacement, I want to estimate the COUNT(DISTINCT V)
in each of the group. However since the bootstrapping is with replacement, some of the element might repeat, and the cardinality of V will be smaller then the original.
For example I could get in one of the groups:
City | Country
--------------
NYC | USA
LA | USA
LON | UK
LON | UK
NYC | USA
and then the estimation of COUNT(DISTINCT Country)
would be 2 instead of 3.
Is there a way to estimate the original value through the values from bootstrapping?