by the screening procedure. This is due to the fact that the screening 

 procedure is able to select variable X12 from lag period 2 where it is 

 stronger than it is in lag period 5, the lag period shown by the Krumbein, 

 Benson and Hempkins (1964) program to be the strongest combination for the 

 given lag period. 



More pronounced differences exist in the kinds of variables chosen 

 by the two programs in the remaining runs, and the variance reduction is 

 perhaps significantly higher for the runs involving (M^) and j£ , where 

 the screening procedure is used. 



SUMMARY REMARKS 



Although the alternative search procedure described in this addendum 

 will not necessarily select the best set of r predictors out of an original 

 set containing P predictors, the screening procedure does allow rapid con- 

 sideration of many more predictors than the Krumbein, Benson and Hempkins 

 (1964) procedure and, in terms of variance reductions, appears to yield 

 comparable predictor equations. Thus, the screening procedure appears to 

 be worthy of application in studies such as this in which large numbers of 

 interrelated predictors are to be considered, and in which the data extend 

 through time. 



ACKNCWLEDGMENT 

 We wish to thank W. C. Krumbein for review of this addendum. 



A-5 



