Abstract
A method for determining climatic change in precipitation data is applied to seasonal data for the United States. This method is an analysis of variance procedure with a Markov chain model used to estimate the within-groups variance. A comparison of the within-groups variance computed with an allowance for dependence to the estimate of the variance computed under the assumption of independence shows that the model with dependence is superior. The statistic which is calculated in the test for climatic change may be significantly inflated if dependence in the data is not modeled. A consequence of this is that climatic change may be inferred when, in fact, it has not occurred. The greatest possibility of this error is in the fall season, and at West Coast stations in all seasons, for the period studied.
