for the wave pole spectrum. 



There were two possible sources of error considered in the stereo data. 

 Even after their removal, there was still a considerable discrepancy. 



The first possible source of error was what is called white noise reading 

 error by Tukey [1949] for the one dimensional case. It can be easily 

 generalized to the two dimensional case. Let 



(11-2) ^jk = ^jk(true) + ^j* + ^k* + 'jl 



where r]--^ is the actual reading, 'Hik/true) ^^ ^^^ reading that would be obtained 

 from the stereo data with the stereo planigraph if there were absolutely no 

 sources of photographic, machine or human error, and e. , ej^"* and e^j^ are 

 random errors. 



More precisely, let e*"' be numbers picked at random from a normal popu- 

 lation with zero mean with an unknown variance and added to every value of a 

 column of 'nik(true) ' ^^^ ^'k. ^® similar number with perhaps a different variance 

 added to every row, and let e-^ be numbers picked at random from still a 

 third different normal population with a zero mesin and a different variance and 



added to the appropriate value of ■n4],/trne) "*" ^i "^ ^k ' '^^^ errors just de- 

 scribed will be referred to as column noise, row noise, and white noise, re- 

 spectively. 



For a more recent and miore readily available reference, see Press and 



Tukey [1957]. 



166 



