The Essentials of Biostatistics for Physicians, Nurses, and Clinicians

7.5 Multiple Regression 113

variability of the size of the counties, and hence all candidate votes
increase together because the increase is primarily due to the larger size
of county.
Using the software package SAS, we looked at three of the possible
multiple regression models. Let N 1 = Gore ’ s total votes in the county,
N 2 = Bush ’ s total, N 3 = Nader ’ s total, and M = Buchanan ’ s total votes
(the dependent variable). The three models are as follows:

1. 
   

   M = β 1 N 1 + β 2 N 2 + β 3 N 3 + α.

   
   


2. 
   

   M = β 2 N 2 + β 3 N 3 + α.

   
   


3. 
   

   M = β 2 N 2 + β 3 N 3 + β 23 N 2 N (^) 3.
   In model 1, the coeffi cient β 1 was not statistically signifi cant. So
   model 1 was dispensed with, and only models 2 and 3 remained under
   consideration. The SAS code used to obtain the results is given in italics
   as follows:
   data fl orida ;
   input county$ gore bush buchanan nader;
   cards;
   alachua 47300 34062 262 3215
   baker 2392 5610 73 53
   ‘
   .
   .
   walton 5637 12176 120 265
   washingtn 2796 4983 88 93
   ;
   run;
   data fl orid2;
   set fl orida;
   if county = ’ palmbch ’ then delete;
   nbinter = nader * bush;