Comments On Silverman's Evaluation Of The Kern River Program


  • Paul J. Smith South Dakota School of Mines and Technology, Rapid City, SD, USA


 am skeptical about Silverman's (2008) approach of taking point estimates of a seeding effect (in this case, from the Kern River program) at face value, as in the interpretation of his Fig.1, and viewing a variation in those estimates with increasing length of record as an indication of a time trend in the seeding effect. A point estimate is just that: An estimate. There is a 10% chance that the "true value" of the effect does not even lay within the 90% confidence interval. I question whether small variations of the point estimate, within a confidence interval that does not substantially narrow with increasing sample size (no plot of the variation of the confidence intervals appears in Silverman 2008, but see Fig. 3 of Silverman 2007 for an example), indicate anything other than simple statistical variations of the point estimate.




Technical Notes and Correspondence