Google
 

Trailing-Edge - PDP-10 Archives - decus_20tap2_198111 - decus/20-0047/cross.txt
There are 6 other files named cross.txt in the archive. Click here to see a list.

	CROSS
	-----

	THE FORTRAN PROGRAM CROSS INCORPORATES MANY OF THE FEATURES OF
	THE DISCRETE BRANCH OF PROJECT IMPRESS DEVELOPED AT DARTMOUTH
	COLLEGE.  USERS FAMILIAR WITH IMPRESS WILL NOTE CERTAIN DIFFER-
	ENCES, SUCH AS A SIMPLIFICATION OF THE COMMAND HIERARCHY AND THE
	FACT THAT VARIABLES AND CATEGORIES ARE NUMBERED INSTEAD OF BEING
	LABELED.

	CROSS IS INTENDED FOR THE ANALYSIS OF DATA COLLECTED WHOLLY OR
	PARTLY BY THE USER, AS, FOR EXAMPLE, A SURVEY CONDUCTED BY A
	SOCIAL SCIENCE CLASS.  OBSERVATIONS OR RESPONSES ARE TO BE CODED
	ACCORDING TO CERTAIN RULES AND THE DATA, PRESUMABLY PUNCHED ON
	CARDS, READ INTO A DATA FILE BY MEANS OF THE FORTRAN PROGRAM
	SURVEY.  PROSPECTIVE USERS OF CROSS SHOULD REFER TO THE INSTRUC-
	TIONS FOR THE PROGRAM SURVEY.

	BEFORE RUNNING THE PROGRAM CROSS TO ANALYZE THE DATA, THE USER
	MAY FIND IT HELPFUL TO OBTAIN A PRELIMINARY TABULATION OF RE-
	SPONSES WITH THE PROGRAM SORTER.


      PROGRAM CROSS
	(CORRELATION OF RESPONSES WITH OPTIONS FOR THE SOCIAL SCIENCES)

      DESCRIPTION
	THIS PROGRAM CORRELATES RESPONSES TO SELECTED QUESTIONS OF A
	QUESTIONNAIRE TO PRODUCE A TABLE OF MARGINAL FREQUENCIES FOR
	ANY VARIABLE OR CROSS-TABULATIONS OF PAIRS OF VARIABLES (WITH
	OR WITHOUT A CONTROL VARIABLE) GIVING FREQUENCIES, PERCENTAGES,
	OR OTHER STATISTICS DESIRED BY THE USER.  AN EXCLUSION ANALYSIS
	IS MADE IN EVERY CASE.  AN ITEM ANALYSIS MAY BE CARRIED OUT ON A
	SELECTED SET OF VARIABLES AND THE RESULTS TABULATED AS A MATRIX
	OF ITEM INTERCORRELATIONS.  IN ADDITION, A MAPPING OPTION ALLOWS
	RESPONSES TO BE GROUPED INTO NEW CATEGORIES, WHILE JOINING AND
	POOLING OPTIONS LET THE USER CONSTRUCT NEW VARIABLES.  MODIFIED
	VERSIONS OF THE DATA FILE MAY BE SAVED FOR USE IN SUBSEQUENT
	RUNS OF THE PROGRAM.  THE USER HAS ACCESS AT ALL TIMES TO EVERY
	VARIABLE IN THE SURVEY (AS MANY AS 128).

	THE DATA ANALYSIS PART OF THIS PROGRAM IS AVAILABLE SEPARATELY
	AS CROSS1.  THE DATA MANIPULATION PART CAN BE RUN SEPARATELY AS
	CROSS2.

      SOURCE
	NORMAN W. JOHNSON, DEPARTMENT OF MATHEMATICS, WHEATON COLLEGE,
	NORTON, MASS.

      INSTRUCTIONS
	THE CODED RESPONSES TO THE QUESTIONNAIRE SHOULD BE READ FROM
	CARDS AND WRITTEN INTO AN ASCII DATA FILE BY THE PROGRAM SURVEY.
	A PRELIMINARY TABULATION OF MARGINAL FREQUENCIES FOR ALL VARI-
	ABLES THE USER INTENDS TO WORK WITH CAN BE OBTAINED BY RUNNING
	THE PROGRAM SORTER.

	THE PROGRAM WILL FIRST REQUEST THE USER TO ENTER THE NAME OF
	THE DATA FILE.  AFTER THIS IS DONE, A DESCRIPTION OF THE SURVEY
	WILL BE TYPED.  THE USER MAY THEN PROCEED TO ANALYZE THE DATA BY
	GIVING ONE OF THE COMMANDS "XTAB" OR "ITEM", MAY REDEFINE OR RE-
	STORE CATEGORIES WITH ONE OF THE COMMANDS "MAP" OR "UNMAP", MAY
	CONSTRUCT OR DELETE VARIABLES WITH ONE OF THE COMMANDS "JOIN",
	"POOL", OR "CUT", MAY TYPE "SAVE" TO PRESERVE THE CURRENT FORM
	OF THE DATA AS A NEW FILE, MAY RESTORE THE ORIGINAL FORM OF THE
	DATA WITH THE COMMAND "RESET", OR MAY TERMINATE THE EXECUTION OF
	THE PROGRAM BY TYPING "STOP".

	COMMANDS THAT TAKE VARIABLES AS ARGUMENTS MAY HAVE THE LIST OF
	VARIABLES ENTERED ON THE SAME LINE AS THE COMMAND.

	THE COMMAND "XTAB" ENABLES THE USER TO OBTAIN MARGINAL FREQUEN-
	CIES FOR ANY ONE VARIABLE OR TO CROSS-TABULATE ANY PAIR OF VARI-
	ABLES, USING A THIRD VARIABLE AS A CONTROL IF DESIRED.  THE
	CONTROL VARIABLE MAY BE RESTRICTED TO ONE CATEGORY BY FOLLOWING
	THE THIRD VARIABLE NUMBER WITH AN EQUALS SIGN (=) AND A CATEGORY
	NUMBER (0 TO 9).  FOLLOWING THE THIRD VARIABLE NUMBER WITH AN
	INEQUALITY SIGN (<>) AND A CATEGORY NUMBER HAS THE EFFECT OF
	EXCLUDING RESPONDENTS IN THAT CATEGORY.  IF TWO OR THREE VARI-
	ABLES ARE SPECIFIED, ONE OR MORE OUTPUT OPTIONS MUST BE SELECTED
	AND THE APPROPRIATE CODE NUMBERS ENTERED.  THE FOLLOWING OPTIONS
	ARE AVAILABLE:

		0--EXCLUSION ANALYSIS ONLY
		1--FREQUENCY OF RESPONSES
		2--PERCENTAGES ACROSS
		3--PERCENTAGES DOWN
		4--PERCENTAGES OF TOTAL
		5--CHI-SQUARE, DEGREES OF FREEDOM, AND PROBABILITY
		6--GOODMAN AND KRUSKAL'S TAU
		7--YULE'S Q OR GOODMAN AND KRUSKAL'S GAMMA
		8--PHI COEFFICIENT OR KENDALL'S TAU
		9--SOMER'S D

	THE COMMAND "ITEM" PRODUCES A MATRIX OF ITEM INTERCORRELATIONS
	(GOODMAN AND KRUSKAL'S GAMMA) FOR AS MANY AS NINE VARIABLES.  A
	VARIABLE PRECEDED BY A MINUS SIGN HAS ITS CATEGORIES REVERSED.
	WHEN THE COMMAND IS GIVEN AS "ITEM PART", A CONTROL VARIABLE MAY
	BE SPECIFIED.  CORRELATIONS MAY BE LIMITED TO A CERTAIN SET OF
	RESPONDENTS BY FOLLOWING THE CONTROL VARIABLE WITH AN EQUALITY
	OR INEQUALITY SIGN AND A CATEGORY NUMBER.

	THE COMMAND "MAP" ENABLES THE USER TO REGROUP THE RESPONSES TO
	ANY QUESTION.  FOR EACH VARIABLE OR BLOCK OF VARIABLES, THE USER
	SPECIFIES HOW MANY NEW CATEGORIES ARE TO BE CREATED, THEN LISTS
	THE CODE NUMBERS (0 TO 13) OF THE RESPONSES TO BE INCLUDED IN
	EACH GROUP.  NUMBERS MAY BE LISTED INDIVIDUALLY, SEPARATED BY
	COMMAS, OR TWO NUMBERS MAY BE JOINED BY A HYPHEN TO FORM A BLOCK
	OF CONSECUTIVE RESPONSES.  BY TYPING "1" OR "KEEP" WHEN THE NUM-
	BER OF NEW CATEGORIES IS REQUESTED, THE USER MAY PRESERVE THE
	CURRENT GROUPING.  TYPING "0" OR "UNMAP" RESTORES THE ORIGINAL
	CODING.  TYPING "-1" OR "REFLECT" REVERSES THE EXISTING CATE-
	GORIES.  TYPING "X" OR "EXCLUDE" ELIMINATES ALL CATEGORIES.  A
	LIST OF THE CURRENT CATEGORIES MAY BE OBTAINED BY TYPING "LIST".

	AFTER EACH MAPPING THE USER MUST INDICATE WHETHER ANY CATEGORIES
	ARE TO BE EXCLUDED.  IF NOT, AN EXCLUSION CODE OF 0 IS ENTERED.
	OTHERWISE, THE EXCLUSION CODE IS 1 IF THE USER DESIRES TO OMIT
	THE HIGHEST NUMBERED CATEGORY, 2 IF THE TWO HIGHEST CATEGORIES
	ARE TO BE OMITTED, ETC.  LIKEWISE, AN EXCLUSION CODE OF -1 MAY
	BE USED TO OMIT THE LOWEST NUMBERED CATEGORY, AND SO ON.  AN EX-
	CLUSION CODE CANNOT LEAVE FEWER THAN TWO CATEGORIES.  HOWEVER,
	THE USER MAY TYPE "X" TO ELIMINATE ALL CATEGORIES.

	THE COMMAND "UNMAP" CANCELS THE EFFECT OF THE CURRENT MAPPING OF
	THE VARIABLE OR VARIABLES SPECIFIED, THUS RESTORING THE ORIGINAL
	CODING OF THE RESPONSES TO EACH QUESTION.  THE ORIGINAL CODING
	IS ALSO RESTORED WHEN THE PROGRAM IS RUN FROM THE BEGINNING,
	UNLESS A MODIFIED VERSION OF THE DATA FILE HAS BEEN SAVED.

	THE COMMAND "JOIN" ALLOWS THE USER TO COMBINE TWO OR THREE VARI-
	ABLES INTO ONE NEW VARIABLE.  ALL COMBINATIONS OF CATEGORIES
	FOR THE GIVEN VARIABLES ARE ARRANGED IN A TABLE, WHICH THE USER
	FILLS IN WITH THE NUMBERS (0 TO 9 WITH 11, 12, AND 13 FOR EX-
	CLUSIONS) TO BE ASSIGNED TO THE CORRESPONDING CATEGORIES OF THE
	NEW VARIABLE.  AS IN THE CASE OF THE "XTAB" COMMAND, WHEN THREE
	VARIABLES ARE LISTED, THE USER MAY SELECT OR REJECT A PARTICULAR
	CATEGORY OF THE THIRD VARIABLE.  THE "JOIN" COMMAND ALSO ALLOWS
	SINGLE VARIABLES TO BE DUPLICATED AND PROVIDES A WAY OF HANDLING
	VARIABLES WITH TWO- OR THREE-DIGIT CATEGORY NUMBERS.

	THE COMMAND "POOL" MAY BE USED TO CONSTRUCT AN INDEX VARIABLE
	FROM A BLOCK OF VARIABLES WITH SCALED RESPONSES, E.G., QUESTIONS
	DESIGNED TO MEASURE RESPONDENTS' ATTITUDES.  THE TOTAL SCORE OF
	EACH RESPONDENT FOR THE BLOCK OF QUESTIONS IS COMPUTED, DIVIDED
	BY THE NUMBER OF QUESTIONS, AND ROUNDED TO THE NEAREST INTEGER.
	THE RESULTING CATEGORY NUMBER PLACES THE RESPONDENT ON A SCALE
	FOR THE INDEX VARIABLE COMPARABLE TO THE ONE EMPLOYED FOR THE
	VARIABLES USED TO CONSTRUCT IT.

	THE COMMAND "CUT" PERMITS THE DELETION OF THE HIGHEST NUMBERED
	VARIABLE OR VARIABLES, EITHER TO MAKE ROOM FOR NEW VARIABLES TO
	BE CREATED BY JOINING OR POOLING OR TO ELIMINATE ONE OR MORE OF
	THOSE MOST RECENTLY CREATED.  THE USER SPECIFIES THE NUMBER OF
	VARIABLES TO BE REMOVED BY ENTERING A DELETION CODE.

	THE COMMAND "SAVE" ENABLES THE USER TO PRESERVE THE CURRENT FORM
	OF THE DATA, INCORPORATING ALL MAPS, JOINS, POOLS, AND CUTS, AS
	A NEW DATA FILE.  THIS FILE MUST BE GIVEN A NAME, DIFFERENT FROM
	THAT OF ANY EXISTING DATA FILE, CONSISTING OF FROM ONE TO FIVE
	CHARACTERS.  THE NAME MAY BE SPECIFIED IN THE "SAVE" COMMAND
	(E.G., "SAVE FILNM") OR SEPARATELY.

	THE COMMAND "RESET" RESTORES THE INITIAL STATE OF THE DATA, THUS
	UNDOING THE EFFECTS OF ALL MAPS, UNMAPS, JOINS, POOLS, AND CUTS.

	AFTER ANY OF THE ABOVE COMMANDS IS CARRIED OUT, A NEW COMMAND
	MAY BE GIVEN.  THE USER MAY SUPPRESS THE PRINTING OF TABLES OR
	ANY OTHER OUTPUT IN ORDER TO GIVE A NEW COMMAND IMMEDIATELY BY
	TYPING <CTRL>O AND PRESSING <RETURN>.  FURTHER EXECUTION OF A
	COMMAND MAY BE HALTED AT ANY BREAK POINT BY TYPING "ABORT".

	WHENEVER THE USER IS EXPECTED TO GIVE A COMMAND OR TO PROVIDE
	CERTAIN INFORMATION NEEDED FOR THE EXECUTION OF A COMMAND, AN
	EXPLANATION OF WHAT IS REQUIRED MAY BE OBTAINED BY TYPING THE
	WORD "EXPLAIN" OR SIMPLY A QUESTION MARK.

	THIS PROGRAM ASSUMES THAT OUTPUT IS TO THE USER TERMINAL.  IF
	OUTPUT TO A FILE ON THE DISK IS DESIRED INSTEAD, THE DISK SHOULD
	BE ASSIGNED LOGICAL UNIT 5 PRIOR TO RUNTIME.

      REMARKS
	THE COMPUTED VALUE OF CHI-SQUARE WITH ONE DEGREE OF FREEDOM
	INCORPORATES A CORRECTION FOR CONTINUITY.  THE CALCULATION OF
	CHI-SQUARE IS ACCOMPANIED BY A WARNING IF THE EXPECTED FREQUENCY
	IS LESS THAN 1 IN ANY CELL OR LESS THAN 5 IN MORE THAN 20 PER
	CENT OF THE CELLS.  KENDALL'S TAU HAS A CORRECTION FOR TIES.

      REFERENCES
	JAMES A. DAVIS, 'ELEMENTARY SURVEY ANALYSIS', PRENTICE-HALL,
	    ENGLEWOOD CLIFFS, N.J., 1971.
	JOHAN GALTUNG, 'THEORY AND METHODS OF SOCIAL RESEARCH', COLUMBIA
	    UNIVERSITY PRESS, NEW YORK, 1967.
	E. TERRENCE JONES, 'CONDUCTING POLITICAL RESEARCH', HARPER &
	    ROW, NEW YORK, 1971.