Statistical Method. 35 



It may appear of little consequence as to whether one subtracts 

 the preliminary first-period values from those which succeed the giving 

 of the dose (the method employed in this paper) or subtracts succeeding 

 values from the preliminary first-period values (the method employed 

 by Dodge and Benedict) . However, by the latter method the resulting 

 signs have, when taken in connection with the discussion in the text, 

 exactly the reverse meaning from that which is usually associated with 

 them. According to Dodge and Benedict (see their page 29) : "If the 

 Av. D. has a minus sign it shows that the measured values are larger 

 as the session progresses. Conversely, if the Av. D. has a positive 

 sign, it shows that on the average the subsequent measurements 

 are less than the normal of the day." It is a real disadvantage for 

 clear understanding of data to have to carry in mind continually this 

 reversal of meaning for + and — which runs through the Dodge and 

 Benedict publication; accordingly, at the risk of some confusion in a com- 

 parison of results, the method represented by the second formula given and 

 illustrated in table 2 has been adopted. Of course the results themselves 

 are not affected by either method of computation, but that method 

 which most facilitates clear understanding and interpretation is to be 

 desired. Wherever data from the Dodge and Benedict report are 

 reprinted for comparison, the liberty is taken to change the signs. 



Dodge and Benedict took the average of the individual differences 

 for all periods of an experimental day as the significant result. The 

 individual periods received no consideration. Our experimental day 

 was 5 hours long as contrasted to the 3 hours used previously. As will 

 be seen later in this paper, the differences produced by the ingestion of 

 alcohol do not always with this subject retain the same sign (plus or 

 minus) throughout the experimental day. For example, a series of 

 plus differences during the first three periods following the dose may 

 change to minus differences during the last two periods of the session. 

 To average together the five differences under these conditions would 

 mask important tendencies and give quite misleading comparisons. 

 The 5-hour experiment was adopted in this repetition series in order 

 that the effect of alcohol might be traced for a longer time after the 

 ingestion of the alcohol. The results are therefore presented as illus- 

 trated by table 2 in a more analytical form and chief attention is given 

 to a comparison by periods following the alcohol. 1 The method used 

 in this report has been to average the differences for homologous periods 

 on normal days and subtract these from a series of such averages for 

 alcohol days in order to obtain an expression of the effect of alcohol for 

 the different periods. This method does not preclude comparison of 

 the averages for individual days, when this is desirable. 



Another item of statistical nature should receive attention. A com- 



1 Dodge and Benedict's report, p. 210, table 39, is the one instance in which these authors have 

 used this form of presentation of their data. 



