This same idea can be applied to other spectral measuring 

 problerns . The question is, "Can the population of spectral curves 

 being studied be analyzed ao though it v/ere being produced with 

 combinations of three or fev/er corr.ponents? " To be more specific, 

 "Can the population of spectral data be matched with linear 

 combinations of three basis curves?" Characteristic vector 

 analysis can be used to ansv/er this question. 



Simonds has described the application of characteristic 

 vector analysis to optical response data and illustrates in detail 

 a procedure for calculating characteristic vectors. For this 

 analysis we first compute the average curve for the population 

 and subtract this average curve from, each of the samiples in the 

 population. An iterative procedure is then used to compute a 

 set of basis curves or vectors. These vectors describe the var- 

 iations of the population- about the average. A small number of 

 characteristic vectors can often describe a complex set of data. 

 In the procedure described by Simonds, one computes the vectors 



31-7 



