no. 3.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 565 



Section 3. — Hollerith analysis. 



Statistical analysis of the principal sample was planned for the Hollerith system of 

 mechanical sorting. In this system it is necessary that the information to be dealt with be 

 coded in numerical terms and then punched into Hollerith record cards. It was decided to 

 code the following items (see reproduction of Hollerith record card, p. 568) : Statistical group 

 number for the principal sample as mentioned above (section 2), camp at which the examination 

 was made, State from which the man was drafted, age, rank, arm of service or whether a 

 recruit, country of nativity in the case of foreign-born men, number of years that foreign-born 

 men have been in the United States, number of years schooling reported, the form of alpha 

 examination, the score on each of the separate tests of the alpha or beta examination (but in 

 the case of men taking both examinations only the score for alpha, since the record card is not 

 large enough to include both sets), total raw score in the alpha examination, total raw score 

 of the first seven tests in the beta examination, the form of the individual examination (i. e., 

 whether performance scale, point scale, or Stanford-Binet scale), the score in individual 

 examination, and finally an index (used later in the statistical combination of results) of 

 the particular combination of examinations taken by the man (whether alpha alone, or alpha 

 with beta, or beta with an individual, etc.). It should be noted that this arrangement makes 

 available considerable information on the separate tests, information which it has, however, 

 not been possible to utilize very fully. The object of giving the alpha scores in the separate 

 tests in preference to the beta scores when both are available was to provide a complete 

 distribution of performances in alpha. The beta group is already highly selected at the 

 upper end by the segregational procedure for dividing between alpha and beta; hence, further 

 distortion due to the elimination of cases at the lower end is not a matter for concern. The 

 total scores for alpha and beta are both given when both examinations were taken. 



The examination scores were coded so as to give the following class intervals: For alpha, 

 5 points; for beta, 5 points; for performance scale, 5 points (an interval, however, which 

 proved too small and which has therefore been doubled — 10 points — in the tables); Stanford- 

 Binet scale, one-half year. 



The principal statistical difficulty encountered in the comparison of these psychological 

 records is the one that occurs when it is desired to compare two groups, each of which contains 

 various numbers of reoords based on the different types of examination. Some theory of 

 combination must be adopted. Alpha, beta, and individual examination need to be expressed 

 in some common system of measurements in order that the comparisons may be simple and 

 brought readily within the range of attention. The difficulties of such a combination and 

 the mathematical development of a scheme are set forth in chapter 2. The scheme of 

 this chapter should have been developed prior to the making of the plan of analysis. It 

 would have been highly desirable, for example, if a rating in terms of an ideal common scale 

 could have been placed upon each record selected and coded on the Hollerith card. Unfor- 

 tunately, however, a shortage of statistical aid not only delayed the completion of this study 

 until after the initiation of the mechanical analysis, but even until after the completion of most 

 of the other studies of this part, which should logically have been based upon the ideal scheme 

 of combination. Failing a common set of terms for the various examinations, it was decided 

 to code the principal sample and to make the Hollerith sortings in such manner that there 

 would be separate tables for each type of examination and also for the cases where one type 

 of examination overlapped another type. The various items given below in the code for 

 "combination of examinations" provide for sortings with respect to these various divisions. 



Rules for coding. — The rules for coding are given herewith in part because they explain in 

 detail the general statements made just above and in part because in the event that these 

 records, now in the Surgeon General's Office, should be used for subsequent research they 

 make plain the details of the information available on the Hollerith cards. 



