Zile,

> Helen, I think you are wrong here. I have a table with about 100.000
> records and index on on column which can have only 3 diferent values.
> Selectivity for thet index show value 0.333, which means that it is computed
> as (1/count of unique values).

both you and Guido seem to imply that the value that you set with SET
STATISTICS is the selectivity of an index.
AFAIK it's not, they are two different figures entirely.
In particular, I don't think that the selectivity of an index is
something that IB keeps track of or use. It can be much more frequently
found in these lists than in IB's documentation. ;-)
Ciao
--
____
_/\/ando