Recomputing index statistics, i.e.
SET STATISTICS INDEX MY_INDEX;
can be achieved by direct updating system table, i.e.

UPDATE RDB$INDICES
SET RDB$STATISTICS = -1
WHERE RDB$INDEX_NAME='MY_INDEX';
COMMIT;

I found this trick years ago, when struggling with optimizer
(the original idea was to convince the optimizer to choose better plans
by imposing my own index statistics.)


Example:
========

CREATE TABLE tab (id INTEGER);
CREATE INDEX ix ON tab(id);

INSERT INTO tab(id) VALUES (10);
INSERT INTO tab(id) VALUES (20);
INSERT INTO tab(id) VALUES (30);

COMMIT;

SELECT RDB$STATISTICS
FROM RDB$INDICES
WHERE RDB$INDEX_NAME='IX';

RDB$STATISTICS
====================
0 <---- old statistics value, not reflecting newly inserted records.


UPDATE RDB$INDICES
SET RDB$INDEX_NAME=RDB$INDEX_NAME <---- just dummy update is enough
WHERE RDB$INDEX_NAME='IX';

SELECT RDB$STATISTICS FROM RDB$INDICES WHERE RDB$INDEX_NAME='IX';

RDB$STATISTICS
====================
0 <---- before commit nothing changes


COMMIT;

SELECT RDB$STATISTICS
FROM RDB$INDICES
WHERE RDB$INDEX_NAME='IX';

RDB$STATISTICS
====================
0.3333333432674408 <---- and after commit the selectivity is recomputed.


SET AUTO OFF;
SET STATISTICS INDEX IX;

SELECT RDB$STATISTICS FROM RDB$INDICES WHERE RDB$INDEX_NAME='IX';
RDB$STATISTICS
====================
-1 <---- just to see what "real" SET STATISTICS is doing before commit.



Ivan
http://www.volny.cz/iprenosil/interbase