who think in such terms. The second set of values entitled 
"Departure from True Values" is simply ten to the power one 
tenth of the numbers in the first set of tables. The third 
set of values is the reciprocal of the second set of values. 
Estimate of sampling error 
Suppose that some fixed power spectrum is chosen. From the 
power spectrum suppose that a section of the function, 7 (t), is 
constructed. And then suppose that a U, from (10.32) is found as 
an estimate of the power in some band by the use of equations 
(10.29), (10.30), (10.31) and (10.32). The true value is known 
from the chosen form of the fixed power spectrum, and the estimate 
is known by the procedures given by Tukey and Hamming [1949]. How 
far off can the estimate be? The answer to the question can be 
given in a probability sense. The estimate is a sample from a 
population of possible samples. That is, many different samples 
could have been taken from many different 7 (t) constructed from 
the same fixed power spectrun. 
For a particular example, suppose that there are ten degrees 
of freedom (i.e. f = 10). Then there is one chance in forty that 
the estimate will be less than 32% of the true value. There is 
one chance in twenty that the estimate will be less than 39% of 
the true value. There are nine chances in ten that the estimate 
will lie between 39% and 180% of the true value. There is one 
chance in twenty that the estimate will exceed the true value by 
180%. Finally there is one chance in forty that the estimate will 
exceed 200% of the true value. The estimate (from the center 
column) will be too low more often than it is too high. There 
- 283 - 
