General Report on Tunny
25W Page 191
(d) Decibanning a run on its own message and correct wheels.
If δ is to be evaluated from the evidence of the message under consideration only, then 
where x* is the double bulge on the correct wheels. This is a compact and convenient notion, though until the correct wheels are known, x* is unknown and must be estimated.
(e) Decibanning a run from its own wheel.
To simplify the calculation it is assumed
(i) that a sufficient approximation to the expected δ is that value of δ whose expected 
is the observed .
(ii) that the distribution of x is normal. This is satisfactory if x << R, a condition satisfied except in key.
It will be convenient, though using the same notation as before, to interpret it more generally by taking Z to be the deviation of a typical double bulge from its mean, 
, so that the variance of Z will be 
. Previously we took δ = 0.
Suppose that the wheel is constructed entirely from the run. Then the score, for a single character is
 |
W10 | |
| and the expected score, x, for the whole wheel is | ||
 |
W11 | |
 |
||
which reduces to  |
W12 | |
where  |
W13 | |
remembering  |
W14 | |
 |
W15 | |
| and eliminating σ, δ from W13, W14, W15 | ||
 |
W16 | |
 |
W17 | |
Back to General Report on Tunny. Contents.