May 28 (Reuters) - One of new Federal Reserve Chairman Kevin Warsh's favorite inflation measures came in cool again on Thursday, offering evidence for his belief inflation is improving and against the view of a growing number of other policymakers that interest rate hikes may be needed to tamp down rising price pressures.

Year-over-year inflation by the Dallas Fed's trimmed mean measure - the best-known of what Warsh referred to in his confirmation hearing as "trimmed averages" for inflation - was 2.3% in April, the Fed bank reported on Wednesday, down from 2.4% in March.

Trouble is, not even the trimmed mean measure's publishers think it's a good yardstick of the underlying inflation trend right now.

"You would want to be cautious on getting too much optimism from the level of the trimmed mean," Dallas Fed economist Tyler Atkinson explained in an interview on Wednesday ahead of the latest data release.

In normal times, he said, the gauge works well to filter out the noise from outlying components, which in April included surging prices for gasoline, airfare and jewelry, and a drop in the prices for poultry, household linens and haircuts. It trims off the fastest-rising prices and fastest-falling prices, leaving a more representative middle set of price changes that typically serves as a good indicator of where inflation is heading.

Items with falling or very slow-rising prices typically outnumber items with steeply rising prices, so Dallas Fed researchers compensate by lopping off more high-inflation items than low-inflation ones.

But lately - because tariffs imposed by President Donald Trump over the last year have pushed up prices for a large chunk of goods - the usual "skew" is reversed. That means that trimming the top 31% and the bottom 24% of items in the index, as the gauge's methodology calls for, ends up pushing the gauge downward, understating true price pressures, Atkinson explained.

This has happened twice before, in the aftermath of the financial crisis and during the post-pandemic inflation surge. Both times, the trimmed mean gave a false signal, suggesting inflation would be cooler than it turned out to be.

By contrast, the measure Fed policymakers have been using for some time to gauge underlying price pressures - the core personal consumption expenditures price index excluding volatile energy and food prices - rose 3.3% in the 12 months through April, the Commerce Department's Bureau of Economic Analysis said on Thursday.

That is the fastest since 2023 and is, as Fed Governor Lisa Cook said on Wednesday, "clearly moving in the wrong direction."

At his confirmation hearing last month, though, Warsh told lawmakers he prefers to take his signal from "trimmed averages" and, as he told Democratic Senator Catherine Cortez Masto, he believes inflation "has improved somewhat in the last year."

Analysts are skeptical.

"We think it is difficult to argue that the disinflation signaled by the trimmed mean is real," wrote Standard Chartered Bank analysts Steve Englander and Dan Pan, noting not just the Dallas Fed measure's statistical properties but also that it historically has not been as good at predicting future inflation as core PCE.

The Dallas Fed has no plans to change its methodology, Atkinson said. If tariff-induced price pressures recede as expected, the problem should cure itself in coming months. Until then, maybe look elsewhere for a guide on the future of inflation, he suggested.

"At the Dallas Fed, we really like the trimmed mean," he said. "But any policymaker would say, you don't just look at a single measure, look at lots and lots of them."

​