AS EACH new economic statistic appears, someone immediately brandishes it as evidence that the president's plans either are or are not working. As the quarrel over the Reagan strategy gets more tense, one elementary question becomes more important: exactly how good are these statistics?

They are pretty good by any technical standard, and there is absolutely no hint of political manipulation. But you need to be aware that there is a lot of wobble in them. The range of uncertainty in the monthly and quarterly figures is rather large, compared with the kind of movements that have become significant in the political quarrel.

American economic policy, and a good deal of American politics, is now being steered by three basic numbers--the unemployment rate, the consumer price index and the gross national product. (The Producer Price Index attracts notice only rarely, when it swings either very high or--as it did last month-- very low.) The tendency of the consumer price index to overstate inflation in the upswings, because of the peculiar way it handles home mortgage interest, is pretty well known. How about unemployment? As it was initially reported, the unemployment rate in December was 8.9 percent of the labor force, in January 8.5 percent and in February 8.8 percent. How do you account for that roller coaster?

Answer: you don't. It is very possible that there was no change at all during the period, or perhaps a slowly and steadily rising trend. Revisions of two-tenths of a point are common. One economist, Lawrence H. Summers, writing in the Brookings Papers, concludes that the rate has to move more than four-tenths of a point before you can be quite sure--before, as a statistician would prefer to say, you can have 95 percent confidence--that the rate has really moved at all. About half of this wobble comes from trouble with the seasonal adjustments; the rest comes from the inevitable imprecision of sampling.

The GNP number, which attempts to track the fluctuations in the economy's entire output, is equally subject to imprecision. Last fall, for example, the government's computers reported that GNP had risen at an annual rate of 1.4 percent for the summer quarter --heartening news, since a lot of people feared that the country was in a recession. But the average revision, between the first report that catches the headlines and the final calculation several years later, is, by an instructive coincidence, also 1.4 percent. It's prudent to remember that the first reported GNP number can be off two or three percentage points. The Bureau of Economic Analysis always offers that reminder on the last page of its GNP press release, but it draws less attention than it deserves.

Governments here and elsewhere have occasionally generated serious trouble by overreactions to economic statistics that later turned out to be misleading. That is most likely to happen in periods like the present one, when things are going badly and anxieties are great. The rule of wisdom, for people who read newspapers as well as for people who make policy, is to pay more attention to general trends over the months than to the latest flash number. An unexpected number may mean that a trend is changing. Then again, as time passes, it may also be the number that gets changed.