I am interested in implementing the (directional) forecasting accuracy test of Henrisksson and Merton (1981) "On Market Timing and Investment Performance II. Statistical Procedures for Evaluating Forecasting Skills" Journal of Business 54, 513-533. Has anyone implemented the test (or is there a procedure) in RATS before?
If my understanding is correct, the test requires just running a regression of a dummy taking the value one when the change in the actual series is positive (zero otherswise) on a constant and another dummy taking the value one when the change in the forecast is positive (and zero otherwise). Then the significance of the slope in the regression is tested (a one sided test). Or is it more complicated than that?