ADEENDUM 



ALTERNATIVE MULT IREGRES SIGN TECHNIQUE FUR 

 OBTAINING PREDICTOR EQUATIONS 



by 



W. Harrison 



U.S. Coast and Geodetic Survey, Washington, D.C. 



and 



N. A. Pore 

 U.S. Weather Bureau, Washington, D.C. 



ABSTRACT 



The data used in the study of interactions in the beach- 

 ocean-atmosphere system are subjected to a multiple-regression 

 screening procedure which is programmed so that all Xs , from 

 all time lags are subjected to a single screening. Examples of 

 predictor equations suggested by the analysis are given for the 

 following predictands: mean longshore current velocity, mean 

 bottom slope and mean grain size in the shoaling-wave zone, and 

 net deposition and net erosion on the lower foreshore during 

 June and July. The screening technique yields comparable or 

 somewhat stronger predictor equations than are obtained in the 

 sequential multiregression analysis, where combinations of Xs 

 are restricted to specific lag periods. 



INTRODUCTION 



The computer program for the least-squares search procedure used in 

 the first part of this report permitted identification of the strongest 

 combinations of 10-14 Xs (predictors) for each Y (predictand) , when the 

 combinations of Xs were restricted to specific lag periods. This scheme 

 was adopted because of a program restriction that limited to 25 the number 

 of Xs that could be studied at any one time, and because of the great 

 numbers of combinations that would be generated if more than a few tens 

 of Xs were considered. 



It is desirable, nevertheless, to have available an extension of the 

 method for searching large numbers of predictors so that each of many Xs 

 may be considered, regardless of its position in time . One such search 

 procedure - called here the screening procedure - identifies "optimum" 

 sets of predictors for given predictands. The data to which the screening 

 procedure will be applied will be identical in every way with those of the 

 first part of this report, to which the Krumbein, Benson, and Hempkins 

 (1964) computer program was applied. 



A-l 



