By rejecting the referendum petition against the domestic partnership law last week, the secretary of state's office saved us all from a repeat campaign on this old and tiresome topic. For that we may be grateful, but we have reason to wonder whether it was rejected the right way.

The petition came extremely close to qualifying for the ballot, and the question is whether in such close cases, the formula being used by the state is good enough, or whether it unfairly works against the citizens trying to bring an issue to a statewide vote.

Their petition against the partnership law needed 55,179 signatures. They turned in more than 60,000. (They claimed 63,000, but the elections office said they actually ended up with only 60,531 raw signatures to check. The office said some of the signatures claimed for this measure were actually for the non-discrimination law the sponsors also intended to refer, and which also failed to qualify.)

Checking all those signatures would take months. So the law provides for statistical sampling instead- two samples in fact, in case the first one falls short.

The first sampling on the domestic-partnership referendum petition came up with a validity rate of 90.7 percent, which would have resulted in about 54,900 signatures, short some 279.

The second yielded a validity rate of 91.63 percent. This would result in some 55,465 valid names, or 286 more than required. Unfortunately for the sponsors, the second sample turned up one duplicated signature. Under the formula, that one duplication caused a reduction in the presumed number of valid signatures of 398. And that, of course, left the proposal short.

The two samples together amounted to slightly more than 5 percent - or one-twentieth - of the number of submitted raw signatures. It would be reasonable to assume that if one-twentieth contained one duplicate, all the signatures would have contained 20 times that many, or 20.

So why penalize the petition signers by almost 400 signatures? The state cites the statistician who devised the formula, a retired Oregon State professor, to this effect: The odds of finding a duplicate in just a sample of just 5 percent are so small that the number actually found has to be multiplied not just by 20 but by 40.

The backers were disappointed, obviously, by how this formula worked out. They have a point. Maybe the procedure could be revised along the lines of what happens in close elections, when the law calls for a recount of every vote.

With enough time - and in referral attempts there is more time than with initiatives - we should count every valid signature if the outcome of mere sampling is very, very close.