center. Similarly for h equal to m, pb h should be given by 
E( Hy) - ECP yy /o)> In addition the formulas given in equation 
(10.37), must be employed in equation (10.32) in order to deter- 
mine the appropriate values for h equal to zero or m. The formulas 
follow from more detailed considerations based upon the fact that 
Q(p) is an even function; i.e., in equation (10.29) summation over 
p equals zero, minus one, etc., to minus m, will result in the 
Same numerical values for Q(-p) as for Q(p). 
Planning the analysis and the work involved 
The point is rapidly being approached where it will be neces- 
sary to use a computing machine, an I.B.M. machine, or an auto- 
correlator such as the one at Woods Hole in order to carry out 
the numerical work indicated in equations (10.29), (10.30), and 
(10.31). The word "work" is chosen advisedly because it will be 
work to do a sufficient number of wave records in order to cali- 
brate the various mechanical electrical wave analyzers now in use. 
This work can be planned. It is possible to decide before- 
hand how reliable the values need to be and how to get these values 
as economically as possible. An example will be given later which 
will show how easily much effort can be wasted. 
The choice of At and the determination of the amount of aliasing 
ee ee ee 
From equations (10.25) and (10.29), the data to be analyzed 
are presented as point values of the original time series at equal 
time intervals. The problem of the choice of At is very important. 
If At is picked too small, too many computations have to be made, 
and the final results often lead to the result that the power 
spectrum is negligible above a certain value of wp. If Nise sis 
= 277 E= 
