General Report on Tunny
23H Page 93
So little evidence is obtained from the 2σ score that the full sigma - a four-wheel run is needed [23C(a)] [R2 p. 76; R5 pp. 35, 54, 97].
(d) Subsequent Runs.
The number of theoretically possible runs is large. If there are two possible runs for the same wheels there is clearly some advantage in combining them: it saves time; the text is longer and if the runs are of similar strength the expected sigma-age is higher; but if, contrary to expectation, one of them is weak, or goes the wrong way, a great deal of evidence is lost. When flogging very hard it is better to keep runs separate and to combine their evidence by the methods of 23J.
Note: if two runs whose texts and proportional bulges are n1δ1; n2δ2, are run together the sigma-age is greater than that of the run n1δ1, if
|
if n2 << n1, this is ½; if n2 = n1, this is
(e) Construction of useful runs.
It is unnecessary to enumerate all runs: consider how such runs may be devised. Suppose that Χ1Χ2 are set, and that it is desired to set Χ3Χ4. Clearly two letters differing only in the fifth impulse will be indistinguishable. Moreover when counting, for example, /T i.e. 1 . 2 . 3 . 4 . , it will be necessary to look at all places where 1 = . 2 = . it is convinient to set forth the 32 letter alphabet thus:
|
| Where 1.2. | a good run is |
|
i.e. 3./1.2. |
| 1.2x |
a weak but useable run is |
|
i.e. 3+4x/1.2x |
| 1x2x | a good run is |
|
i.e. 3+4x/1x2x |
|
or if 8' are
numerous |
|
not 3.4./1x2x | |
| 1x2. | a good run is |
|
i.e. 3x/1x2. |
|
a sometime usable run is |
|
i.e. not 3.4x/1x2. |
3+4x/1.2x and 3+4x/1x2x could be combined into 3+4x/2x, but the run 3+4x/1.2x is so much weaker than the other that this would be unwise.
Not infrequently 3x/1x2. and 3./1.2. can profitably be combined as 3+/1.2.
Back to General Report on Tunny. Contents.