The Federal Reserve says it will depend on straightforward data analysis in deciding when to raise U.S. interest rates.

But whose data will it depend on?

The large half a percentage point revision to estimated U.S. first-quarter gross domestic product issued on Wednesday by the Bureau of Economic Analysis nearly erased a contraction in output that had unnerved policymakers.

In particular it found that exporters and consumers did better than initially thought. The BEA now says GDP shrank only 0.2 percent over the first three months of the year, not 0.7 percent.

The revision may not change the Fed's policy path, but that's only because the consensus about how to measure U.S. economic growth has begun to fracture.

There are competing views among Fed, BEA and private sector economists on the statistical methodology, particularly over such issues as seasonal adjustment, that have led to mistrust of the data.

As a result, the Fed's policy of "data dependence" may turn out to involve a good bit of guesstimating, with Fed officials each playing their own hunch about where the ­economy really stands.

"We don't measure GDP particularly well over short periods of time. It is very difficult. If you dig into it, you will see there are just a lot of extremely difficult calls," Fed Governor Jerome Powell said on Tuesday.

Researchers at the Federal Reserve board in Washington, D.C., the New York Fed, and the San Francisco Fed have published competing papers in recent weeks looking at whether the BEA's methods for seasonal adjustment are creating consistent errors in the GDP estimates.

The answer: maybe. They each had different conclusions. The BEA has agreed there may be a problem and is studying how to revise its seasonal adjustment methods.