@fredbenenson @johnmyleswhite ... and the experiment name makes a convenient salt.
@johnmyleswhite @seanjtaylor @fredbenenson yes, good point. I would still say you want a hash function, just that a dumb fast one is fine.
-
-
@avibryant@johnmyleswhite@seanjtaylor@fredbenenson if you are at the point of wanting to optimise, say, murmurhash then congratulations! -
@mat_kelcey@avibryant@seanjtaylor@fredbenenson Why do you need to better than MurmurHash when you only need 1 bit of entropy? -
@johnmyleswhite@mat_kelcey@avibryant@seanjtaylor@fredbenenson that's not true: most people are in multiple experiments. -
@josh_wills@mat_kelcey@avibryant@seanjtaylor@fredbenenson Use Sean's strategy and add the same constant offset for each experiment? -
@johnmyleswhite@mat_kelcey@avibryant@seanjtaylor@fredbenenson I'm assuming everyone has read http://research.google.com/pubs/pub36500.html … ? -
@josh_wills@mat_kelcey@avibryant@seanjtaylor@fredbenenson I never had. Thanks for the link. -
@johnmyleswhite@josh_wills@mat_kelcey@avibryant@seanjtaylor@fredbenenson Also look up Kohavi's papers on AB testing.
End of conversation
New conversation -
-
-
@avibryant@johnmyleswhite@fredbenenson Here's how dumb: hash function for our Science paper was mod(comment_id + time_stamp, 100) <= 4 -
@seanjtaylor@avibryant@fredbenenson That worked. Would have thought 100 was too composite for modular values to be well dispersed.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.