lxxii 



Tables for Statisticians and Biometricians [XLVIII 



We must interpolate between these two series for n = 79, that is we must take 

 0'42 times the first series and 0'58 times the second series. The results are 

 given below, and set against the direct calculation from formula (lxxx), using 

 Table XLIX. 



The interpolation does not give a result very close to the actual series. For 

 example, not more than three syphilitics might be anticipated in 70 °/ of samples 

 of 40 by the interpolated series ; actually not more than 3 are to be expected 

 in 75 % °f samples. At the same time the result is much better than the 

 normal curve theory provides. In the latter case we have 



Hence 



Mean = 40 x 7 4 5 = 2 025, 

 Standard-Deviation = V40 x ^ x J| 

 (3-5 -2-025)/l-387 = 1-064 



1-387. 



and by Table II this value of x corresponds to |(1 + a) = -86, i.e. in 86 % per cent, 

 of samples of 40, we should have not more than 3 cases. It will be seen therefore 

 that (i) the values at the latter end of the Table are not close enough to obtain 

 very accurate results by interpolation, but (ii) that the Gaussian gives a still 

 poorer approximation. 



Illustration (ii). Of 10 patients subjected by a first surgeon to a given 

 operation only one dies. A second surgeon in performing the same operation 

 on 7 patients, presumably equally affected, loses 4 cases. Would it be reason- 

 able to assume the second surgeon had inferior operative skill ? 



