26 



UNITED STATES NATIONAL MUSEUM BULLETIN 285 



twice the number of output pages. These cards 

 must be present even if they are blank. 



20. ( 1 ) : Numbers of cards to be read in for each of 



8 variable formats (Instructions 21-28). 

 Punch in fields of 3 in card columns 1-24. 

 The number of cards for each format must be 

 at least one and cannot exceed 5, with the 

 exception of the eighth format, for which up 

 to 8 cards may be present. 



21. (1-5): The format, in FORTRAN II, for the read- 



ing in of data cards. The format must begin 

 with a left-parenthesis and end with a right- 

 parenthesis. 



If the control described in Instruction 7c 

 is greater than zero, the following applies: 



Make provision at the beginning of each 

 data card (or each group of cards if a speci- 

 men is represented by more than one card) 

 for reading in the signals which will determine 

 whether certain variables (the column num- 

 bers of which were specified in Instruction 10) 

 are to be divided ( — 1 ) or not divided ( + 1 ) 

 by a constant, the value of which is specified 

 in Instruction 34. 



There can be no more than 5 signals, and 

 the signals are to be read in as fixed-point 

 (integer) variables by means of an "I" specifi- 

 cation as shown in the format listed with the 

 controls for the sample output. All other 

 variables on the data card must be read in 

 as floating-point variables by means of an 

 "F" or "E" specification. 



In addition to the example given by the 

 listing of control cards for the sample input, 

 the following is given : 



A data card which is punched, beginning in 

 the first card column, 4- 1-1-1001760321947, 

 and which is read in by the format (312, 

 F3.0, 2F3.1, F3.3), will be read in as the 

 following values: 4-1, -1, -1, 1., 76.0, 32.1, 

 and .947. 



For further instruction on format specifi- 

 cation in FORTRAN II, particularly with 

 regard to Hollerith "H" specifications and 

 carriage control, the reader is referred to the 

 FORTRAN General Information Manual 

 (IBM, 1961) or to McCracken (1961). 



22. (1-5): The format, in FORTRAN II, for the 



punching out of the univariate output array. 

 The order in which the variables are punched 



out is the same as the order of the subarrays 

 shown in Figure 1. 



This is also the format for the reading in of 

 maxima for the variables in the univariate 

 output array (Instruction 29) and for the 

 reading in of log minima for the variables 

 which will be transformed to logs by Instruc- 

 tion 18d. 



23. (1-5) : The format, in FORTRAN II, for punching 



out the bivariate array. 



24. (1-5) : The format, in FORTRAN II, for printing 



the univariate output array. The format is 

 repeatedly used for each row of the array 

 throughout all segments, so that the maximum 

 number of decimal places which will be 

 required by any one variable should be used 

 for all variables. The same applies to Instruc- 

 tion 25. 



25. (1-5): The format, in FORTRAN II, for the print- 



ing of the bivariate array. (See Instruction 

 24.) 



26. (1-5) : The format, in FORTRAN II, for printing 



the univariate statistics beneath each table of 

 the univariate array. Column alignment is 

 achieved by matching the field specifications 

 to those of Instruction 24. Note in the listing 

 of the sample input that an "I" (integer) 

 specification must be used for the row of 

 numbers of non-zero values. 



27. (1-5) : The format, in FORTRAN II, for printing 



the univariate statistics beneath each table of 

 the bivariate array, again with "I" specifica- 

 tions for the row of numbers of non-zero 

 values. 



28. (1-5) : The format, in FORTRAN II, for printing 



the bivariate statistics beneath each segment 

 of the bivariate array. 



29. (1- ): Card(s) containing the anticipated maximum 



values of all the variables in the univariate 

 output array of this run or any future run 

 analyzing the same variables. The maxima 

 are hypothetical and are specified in order to 

 determine the calibration of the axes of the 

 machine-plotted bivariate scatter diagrams, 

 which in turn allows the computation of 

 slopes and intercepts in terms of the scale- 

 units of the diagrams. (See Instructions 11 

 and 12 in the description of input data for 

 VPLOT.) In this manner, the calibration is 

 kept uniform throughout all runs so that the 



