One reason why it’s a good idea to avoid spending too much time analyzing month-by-month changes in unemployment numbers is the fact that those who compile the data build in seasonal adjustments that can skew actual results.

The latest “Up and Down Wall Street” column from Randall Forsyth of Barron’s highlights the latest instance in which seasonal adjustments could warp our assessment of the nation’s economic progress.

The White House was quick to trumpet the jobs numbers as proof of its success in its first 50 days. (Candidate Trump had charged that the employment data were phony and that joblessness actually exceeded 40%.) The Lindsey Group’s Peter Boockvar offers a more dispassionate view: While Trump’s victory boosted optimism, the jobs gains also reflect “how frozen businesses likely were going into the election” and the subsequent thaw. Reflecting that earlier chill, nonfarm payrolls averaged monthly growth of 148,000 from October through December.

Increasing warmth, literally instead of figuratively, was reflected in the jobs report for last month, which was the balmiest February since 1954. That heavily skewed the seasonally adjusted numbers enough for David Rosenberg, Gluskin Sheff’s chief economist and strategist, to question them in a report titled “Fake Data?”

So far from the norm was the mercury last month—an average of 41.2 degrees Fahrenheit versus the normal 34.5 degrees—that he advised clients to “treat all the February data with a huge grain of salt (obviously not the salt used to melt snow off the driveway, since there isn’t any).”

The seasonal adjustments are supposed to compensate for the usual deep freeze of February. After applying the fudge factor, construction jobs supposedly jumped by 58,000—triple the norm of the past year, and in a sector that’s slowing, Rosenberg adds. At the same time, only 184,000 folks couldn’t make it to work because of bad weather—half the February norm of 365,000. If a more normal number of workers were stuck home for bad weather during the survey week, the headline payrolls gain would be a tepid 55,000.

Those inconvenient facts aside, the data are sufficiently strong for the FOMC to go ahead with a 25-basis-point hike on Wednesday, a move to which the federal-funds futures market has given a 96% probability.