Abstract
The Climax I and II wintertime orographic cloud seeding experiments have recently been reanalyzed (Mielke et al., 1981c). The primary inference technique of this recent reanalysis involved (i) target-control residual data resulting from a simple linear model fitted by least squares and (ii) analyses based on classical linear rank statistics. The respective problems associated with this inference technique are (i) the residual data are highly dependent on a few very large values due to the model being fit by least squares and (ii) the complex non-Euclidean geometry underlying classical linear rank tests. The purpose of this article is to describe and apply a new inference technique which resolves both of these problems. This new inference technique involves residual data resulting from a median regression line and analyses based on recently developed rank tests associated with multi-response permutation procedures (MRPP). Application of this new inference to specific meteorological partitions of the Climax I and II experiments indicates that the evidence for a seeding effect is a little stronger (i.e., smaller P values) with the new technique than with the old technique for the warm Climax II 500 mb temperature partition and a Climax I 700 mb wind velocity partition, a little weaker for the warm Climax II 700 mb equivalent potential temperature partition, and about the same for the other meteorological partitions examined.
