HASSLE STAT PACK
                               FORTRAN IV - STAT

                                  LARS PALMER
                                   AB HASSLE
                                     FACK
                               431 20 MOLNDAL 1
                                    SWEDEN




































                               HASSLE STAT PACK
                                  F IV - STAT




A)GENERAL 


THE COMPONENTS OF THE TOTAL PACKAGE ARE:  

1)STAT          DESCRIBED HERE
2)OUTLAY        CREATES AN INPUT FILE TO STAT
3)KSORT         REARRANGES ROWS AND COLUMNS IN THE FILE
4)OUTCHK        CHECKS ON UNFORMATTED FILE FOR ERRORS
                NOT ON DECUS TAPE
5)BIGSTA        HANDLES >10 COLUMNS
                NOT ON DECUS TAPE


OPERATING ON OUTLAY FILES ARE  ALSO  SEVERAL  SPECIFIC  PROGRAMS  E.G.
LESQ THAT DOES A NON LINEAR LEAST SQUARES CURVE FIT.  




B)IMPLEMENTATION


AS OF TODAYS DATE THE FOLLOWING ROUTINES ARE IMPLEMENTED:

NUMBER  PERFORMS
-1      LIST
0       NEW DATA FROM SAME FILE
1       MEANS ETC
2       T-TESTS
3       REGRESSION LINES
4       TYPE 2 REGRESSION
5       CORRELATION MATRIX
6       ANOVA ONE SIDED
7       SHEFFE CONTRASTS
8       ANOVA TWO SIDED
13      WILCOXON
14      MANN-WHITNEY
15      KRUSKAL-WALLIS
16      SPEARMAN RANK
17      FRIEDMAN
18      NUMERIC INTEGRATION ON X/Y
19      MEDIANS AND RANK OF COLUMN
20      CROSSTABULATION
21      SCATTER PLOT
22      VARIOUS CONVERSIONS
48      OUTPUT A COMMENT
49      FLAG-SETTING ROUTINE
50      EXIT TO MONITOR






                             F IV - GENERAL-PAGE 1


                               HASSLE STAT PACK
                                  F IV - STAT





C)INDATA AND FILE UNIT NUMBERS 


THE FOLLOWING INDATA FILES ARE USED:  

 UNIT   CONTAINS 

 6      THE DATA - IN OUTLAY FORMAT - TO BE ANALYZED 
 5       A  PARAMETER  FILE CONTAINING THE REQUEST FOR THE ANALYSES TO
        BE PERFORMED.  FORMAT SEE BELOW (D).  


 OUTFILES 

 UNIT   CONTAINS 

 3      THE RESULTS NORMALLY TO LINE PRINTER.  IF NO LINE  PRINTER  IS
        AVAILABLE REASSIGN.  
 LDEVE  ERROR INFORMATION. 
 LDEVL  TRACE OF PROGRAM FLOW.  

 LDEVE AND LDEVL ARE FURTHER SPECIFIED UNDER CARD 49.  
































                             F IV - INDATA-PAGE 2


                               HASSLE STAT PACK
                                  F IV - STAT


D)THE PARAMETER FILE 


NOTE  THAT AT PROGRAM START THE PROGRAM READS THE PARAMETER FILE AS IF
IT HAD JUST SEEN A 0 REQUEST CARD.  

THE REQUEST CARDS  HAVE THE FOLLOWING FORMAT:  
CT(CARD TYPE),ARG1(ARGUMENT 1),ARG2,....  

SOME OF THE REQUEST CARDS MUST BE FOLLOWED BY  PARAMETER CARDS.  
THE MOST COMMONLY USED FORMAT IS:  

CT, KOL1, KOL2,IOPT, WITH KOL1, KOL2 BEING 2 COLS TO BE  COMPARED  AND
IOPT AN ANALYSIS SPECIFIC PARAMETER.  

CARD TYPE      ARGUMENTS    PARAMETER CARDS
-1               0-1            1
 0               0              1(-4) GROUPS
                                +1    COLUMNS
 1               0-2
 2               3
 3               2-3
 4               2-3            0-1
 5               0
 6               0
 7               0-1            0-1
 8               0
13               2
14               2
15               0
16               2
17               0-1
18               2-3
19               0-1
20               2-3
21               2-3
22               1-3
48               0-1            AS SPECIFIED BY ARGUMENT.
49               0-4
50               0

        ****COMMENTS***  

 CARD TYPE 

 -1     THE REQUEST CARD -1 ASKS FOR A LISTING OF THE INPUT DATA.  

        IF  ARG1=0 IT MUST BE FOLLOWED BY A CARD INDICATING THE FORMAT
        FOR THE LISTING.  THIS  MUST  BE  A  STANDARD  FORTRAN  FORMAT
        NOTATION  WITH  AS  MANY  FORMAT  DESCRIPTIONS  AS  THERE  ARE
        COLUMNS IN THE FILE.  IT SHOULD CONTAIN  1H  ,  AS  ITS  FIRST
        ELEMENT.  





                             FIV-PARAMETERS-PAGE 3


                               HASSLE STAT PACK
                                  F IV - STAT


 CARD TYPE 

        IF ARG1 <> 0 ONE OF THREE INTERNAL FORMATS IS USED:  

        -1,1    F10.1 
        -1,2    F10.4 
        -1,3    G12.4 

        ****MISSING DATA**** 
        ALL  ROUTINES  IN THE PROGRAM TREAT A 0-VALUE AS MISSING DATA.
        (NONMISSING DATA = 0 CAN BE SET AS E.G. 1E-25).   FORMAT  -1,3
        IS  NOT  SUITABLE  FOR  MISSING  DATA,  BUT IF THERE ARE ANY A
        SPECIFICATION ERROR WILL BE GENERATED.    

 0      READS NEW DATA.   THE PARAMETER FILE ALWAYS  BEGINS  WITH  THE
        PARAMETER CARD  FOLLOWING THE 0 CARD.   

        ****PARAMETER CARD*** 
        1(-4)    NUMBER  OF  GROUPS  TO  BE  READ  INTO CORE AND THEIR
                NUMBERS.  THE  GROUP  NUMBERS  ARE  NOT  PRESERVED  IN
                CORE. THEY ARE TREATED AS ONE GROUP.  
                TWO SPECIAL FORMS OF THIS INPUT ARE AVAILABLE:  

                1)IF  ONLY  THE  NUMBER  OF  GROUPS IS SPECIFIED IT IS
                TAKEN TO MEAN THE FIRST N SEQUENTIAL GROUPS 
                I.E. 
                5 
                IS INTERPRETED EQUIVALENT TO 
                5,1,2,3,4,5 

                NOTE: IF N>24 USE SEVERAL CARDS SEE # 

                2)IF THE FIRST CARD IN THE INPUT FILE IS  NEGATIVE  IT
                IMPLIES  THAT  THE  ANALYSES  FROM  HERE TO THE NEXT 0
                CARD, OR TO THE 50 CARD IF NO 0 CARD  INTERVENES,  ARE
                TO BE DONE FOR ALL GROUPS IN THE INPUT FILE 
                I.E.  
                -1 
                5 
                1 
                5 
                0 
                IMPLIES  'ON  ALL  GROUPS  IN INPUT FILE GENERATE MEAN
                AND CORRELATION MATRIX FOR THE FIRST 5  COLUMNS'.   IF
                UNIT  5  IS  ASSIGNED TO UNIT 4 (E.G. THE REQUESTS ARE
                GIVEN FROM THE CONSOLE)  THIS  OPTION  SHOULD  NOT  BE
                USED.  


             #   THESE CARDS ARE IN 25I3 FORMATS AND MUST IF NECESSARY
                BE CARRIED OVER TO ANOTHER CARD.   






                             FIV-PARAMETERS-PAGE 4


                               HASSLE STAT PACK
                                  F IV - STAT


 CARD TYPE 

        NEXT    NUMBER OF COLUMNS TO BE USED AND THEIR  NUMBERS.   THE
                COLUMNS  CAN  BE  REORGANIZED  IN  CORE BY PLACING THE
                REQUEST IN A NON-SEQUENTIAL MANNER.  SPECIAL  FORM  1)
                AS DESCRIBED UNDER ABOVE ALSO APPLIES HERE.  

                IF  NUMBER  OF  COL IS STATED > THAN NUMBER OF COLUMNS
                IN FILE IT IS ADJUSTED  TO  THE  LATTER.IF  A  TOTALLY
                EMPTY  COLUMN  IS  READ THE COLUMN COUNT IS REDUCED BY
                ANALYSIS NO 1.THEREFORE IT IS  POSSIBLE  TO  ROUTINELY
                SPECIFY  E.G.  10 AS NUMBER OF COLUMNS IF ALL EXISTING
                COLUMNS ARE WANTED.  

 1      THE TREATMENT OF THE ARGUMENTS IS:  
        IF ARG1 AND  ARG2  BOTH  =  0  CALCULATE  MEANS  AND  STANDARD
        ERRORS.   

        IF  ARG1=15  CALCULATE  MEANS  ONLY  ON ROWS THAT ARE COMPLETE
        I.E. CONTAIN NO MISSING DATA SEE 7.   

        IF ARG1<=NK AND ARG2=0 PRINT SUMS OF SQUARES OF COL(ARG1).  

        IF ARG1 AND ARG2  BOTH   <=  NK  PRINT  SUMS  OF  SQUARES  AND
        CROSSUMS  FOR  ALL  COMPLETE  PAIRS. IF ONE DATA IS MISSING SO
        ARE BOTH SKIPPED.   

 2      3 TYPES OF T- TESTS ARE IN THE PACKAGE.  

        CALLING:        2,K1,K2,N 
        N IS TYPE OF TEST AND K1 K2 THE 2 COLUMNS TO COMPARE.   

        N=1     STANDARD FORMULA. SIGMA PRESUMED EQUAL.  
        N=2     SIGMA NOT PRESUMED EQUAL.  
        N=3     PAIRED VALUES.  

 3      REGRESSION LINE - CALLING 3,K1,K2 

 4      TYPE II REGRESSION (BARTLETTS 3 GROUP) REF  S&R  PAGE  480.THE
        NORMAL  REGRESSION LINE PRESUMES X TO BE WITHOUT ERROR.  IF SO
        IS NOT THE CASE ANALYSIS 4 IS TO BE USED.  
        CALLING  4,KOLX,KOLY,IOPT       OPTIONAL 
        THE LINE (Y=A+BX) IS CALCULATED AND  CONFIDENCE  LIMITS  TO  B
        ARE PRINTED.  
        IF IOPT = 0 THE VALUES P=0.05,0.01,0.001 ARE USED.  
        IF  IOPT>0  -  IOPT(<4)  P VALUES MUST BE GIVEN.  THESE ARE IN
        THE FORM 0<P<0.5.  










                             FIV-PARAMETERS-PAGE 5


                               HASSLE STAT PACK
                                  F IV - STAT


 CARD TYPE 

 5      CORRELATION. NO ARGUMENTS.  
        NUMBERS OF PAIRS IN EACH COMBINATION IS PRINTED  IF  THEY  ARE
        UNEQUAL.   

 6      ONEWAY ANALYSIS OF VARIANCE(ANOVA).  

 8      TWOWAY D:O      (SEE ALSO UNDER 7).   

 7       SHEFFES  CONTRASTS.  TO PERFORM THIS ANALYSIS IT IS NECESSARY
        TO HAVE SOME SPECIAL INFORMATION:  

        1)MEANS MUST BE CALCULATED 
        2)THE MEAN SQUARE MUST BE CALCULATED 
        3)THE PROGRAM MUST BE SUPPLIED WITH THE RELEVANT P-VALUES.  

        THEREFORE THE FOLLOWING IS REQUIRED WHEN DOING SHEFFE:  

                1)ANALYSIS 1 MUST BE PERFORMED. IF  THE  ANOVA  TO  BE
                USED  IS  A TWOWAY ANOVA AND THERE IS MISSING DATA THE
                MEANS MUST ONLY BE CALCULATED ON FULL  ROWS  (ANY  ROW
                WITH  MISSING  DATA IS EXCLUDED). THIS IS SIGNALLED BY
                REQUESTING ANALYSIS 1,15.  

                2)THE RELEVANT ANOVA MUST BE REQUESTED.  IF  THERE  IS
                NO  MISSING  DATA  ANALYSIS 1 OR 1,15 MAY BE USED WITH
                EITHER ANOVA BUT  IF  THERE  IS  MISSING  DATA  IT  IS
                IMPERATIVE   TO   USE   THE   PAIRS   STAT6-STAT1   OR
                STAT8-STAT1,15.  

                3)FINALLY ANALYSIS 7 IS REQUESTED  AND  THE  P  VALUES
                HAVE  TO  BE  SUPPLIED  TO  THE  PROGRAM IN ONE OF TWO
                WAYS:  

                A)IF ARG IS NON-ZERO THE NEXT CARD SHOULD  CONTAIN  AS
                MANY  P-VALUES  AS  ARG  (MAX=3)   SEPARATED BY COMMAS
                WITH THE LARGEST FIRST.  THE P  VALUES  SHOULD  BE  IN
                THE  FORM  0<P<0.5  OR  A  SPECIFICATION ERROR WILL BE
                GENERATED.  

                B)IF ARG IS  ZERO:   THE  PROGRAM  USES  THE  INTERNAL
                DEFAULT VALUES OF 0.05,0.01,0.001.  

                THE  OUTPUT  IS  IN  THE  FORM  OF  TWO MATRICES - ONE
                SHOWING THOSE CONTRASTS  EXCEEDING  THE  GIVEN  LIMITS
                AND  ONE  SHOWING  THE  EXACT  PROBABILITIES  FOR  THE
                CONTRASTS.  

 13     WILCOXON        CALLING 13,K1,K2 

 14     MANN WHITNEY    CALLING 14,K1,K2 





                             FIV-PARAMETERS-PAGE 6


                               HASSLE STAT PACK
                                  F IV - STAT


 CARD TYPE 

 15     KRUSKALL WALLIS. IF THERE IS NO MISSING DATA  OR  IF  15,1  IS
        REQUESTED  THE TABLE OF RANKDIFFERENCES IS ALSO OUTPUT (IN THE
        LATTER  CASE  BY  IGNORING  ANY  ROW  IN  WHICH  ANY  DATA  IS
        MISSING).  
        SEE COLQUHOUN.  

 16     SPEARMAN RANK   CALLING 16,K1,K2  

 17     FRIEDMAN ANALYSIS. THE RANK DIFFERENCES ARE CALCULATED.    

 18     NUMERICAL INTERGRATION BY OVERLAPPING PARABOLAS.  
        K1 IS USED AS X AND K2 AS Y.  
        IF ARG1=0 ONLY FINAL AREA 
        IF ARG1=1 ALSO TABULATES ALL PARTIAL AREAS 
        THE  AREAS  ARE NORMALLY CALCULATED BY A METHOD OF OVERLAPPING
        PARABOLAS.  ALTHOUGH A VERY GOOD METHOD,  IT  FAILS  MISERABLY
        ON  SOME KINDS OF EXPERIMENTAL DATA.  A CHECK ON THIS HAS BEEN
        INCORPORATED.    AREAS  ARE  ALSO  CALCULATED  BY  A   TRAPEZE
        FORMULA AND THE TWO AREAS ARE COMPARED.   

 19     MEDIANS AND RANK OF COLUMN 
        IOPT=0 
                MEDIANS  ARE CALCULATED ONCE ON EACH COLUMN CONSIDERED
                SEPARATELY AND IF  MISSING  DATA  ALSO  ONCE  ON  EACH
                COLUMN  SKIPPING  ANY  DATA  IN  A  ROW WHERE THERE IS
                MISSING DATA. THIS IS EQUIVALENT TO ANALYSIS 1,15.  

        0<IOPT<=NK 
                COLUMN IOPT IS RANKED AND THE RANKS  ARE  PRINTED  FOR
                EACH  OBSERVATION.   ALSO  PRINTED ARE THE TWO TIESUMS
                (USED FOR NONPARAMETRIC  ANALYSES).  
                TIESUM1=SUM(CT**3-CT)/12 
                TIESUM2=SUM(CT*(CT-11)/2) 

        FOR USAGE CHECK E.G. SIEGEL  

 20      CROSSTABULATION  OF  2  COLUMNS   OR   TABULATION   ("NUMERIC
        HISTOGRAM")  OF  ONE.  THE LATTER SPECIFIED BY KOL1=KOL2.  THE
        NUMBER OF INTERVALS CAN BE 3 TO 10 WITH 5 AS THE DEFAULT.  

 21     SCATTER PLOT. KOL1 IS  X-AXIS  AND  KOL2  Y-AXIS.   X-AXIS  IS
        ALWAYS  DIVIDED  INTO  50 STEPS Y-AXIS CAN BE DIVIDED  INTO 10
        TO 50 STEPS  SPECIFIED  BY  IOPT.  DEFAULT  IS  25  STEPS.  AS
        LOGARITHMIC   PLOTS   ARE  COMMON  A  SPECIAL  CALL  HAS  BEEN
        INCORPORATED TO GENERATE E-LOGS OF THE X-AXIS.  THIS  IS  DONE
        BY  ADDING 100 TO THE  IOPT ARGUMENT I.E. 22,1,2,110 REGUEST A
        PLOT WITH 10 STEPS ON  THE  VERTICAL  AXIS  AND  LOGS  ON  THE
        HORIZONTAL.      ALSO  NOTE  THAT  LOGS  CAN  BE  GENERATED BY
        OPTION 22.   






                             FIV-PARAMETERS-PAGE 7


                               HASSLE STAT PACK
                                  F IV - STAT


 CARD TYPE 

 22     TRANSFORMATIONS ON THE DATA IN CORE.  THE ORIGINAL  DATA  FILE
        IS  NOT  CHANGED. WHEN SO IS REQUESTED USE KSORT.  ANALYSIS 22
        ONLY  HANDLES  TRANSFORMATIONS  WHERE  ONE  INPUT  COLUMN   IS
        INVOLVED.  IF  MORE  ARE  INVOLVED  E.G. KOL1 IN % OF KOL2 THE
        TRANSFORMATIONS HAVE TO BE DONE VIA KSORT.  

        THE CALLS ARE:  

        22,KOL1         SORTS THE DATA MATRIX BY KOL1. MISSING  (ZERO)
                        VALUES ARE ALWAYS PLACED LAST.  

        22,-KOL1        SORTS BY KOL1 IN DESCENDING ORDER.  

        22,NK,KOL1,-1    REDUCES  DATA MATRIX FROM NK COLUMNS (MUST BE
                        GIVEN)TO KOL1 COLUMNS. ONLY VALID IF  THE  ATT
                        SWITCH  (SEE THE 49 CARD) IS SET.  CAN BE USED
                        INSTEAD OF  RE-READING  THE  DATA  WITH  FEWER
                        COLUMNS  WHEN  THE DATA IN CORE IS TOO BIG FOR
                        A REQUIRED ANALYSIS.  

        22,KOL1,KOL2,CODE  SETS KOL2 TO CODE (KOL1).  

        THE CODES ARE:  
        1       SQUARE ROOT 
        2       EXP(E**X) 
        3       NATURAL LOGS 
        4       LOGS 
        5       SQUARES 
        6       EQUAL I.E. COPIES KOL1 INTO KOL2 

 48     OUTPUT A COMMENT. SOMETIMES IT IS USEFUL TO BE ABLE TO PASS  A
        COMMENT  TO  THE  OUTPUT FILE ON UNIT 3, E.G. WHEN  A STANDARD
        PARAMETER FILE OFTEN IS USED ON DIFFERENT DATA.   STATING  48,
        IOPT  WILL  PROMPT  THE PROGRAM TO ASK FOR IOPT CARDS FROM THE
        INPUT FILE (UNIT 5).IF THE  ATT  FLAG  IS  SET  (SEE  49)  THE
        PROGRAM  PROMPTS  THE  OPERATOR  TO  GIVE  THE COMMENT LINE BY
        OUTPUTTING A NUMBER TO THE TTY (UNIT 0) SO AS TO MINIMIZE  THE
        RISK OF GETTING A LETTER IN A NUMERIC FIELD ON MANUAL INPUT.  

 49      FLAG  SETTING  ROUTINE. THE PROGRAM USES 3 FLAGS AND 2 DEVICE
        NUMBERS TO  SCHEDULE CERTAIN PARTS  OF  THE  INFORMATION.THESE
        ARE:  

        LIST     IF TRUE - PRINT THE TRACE INFORMATION.  USEFUL IF THE
                PROGRAM DOES A LONG UNATTENDED RUN.  

        ATT     IF FALSE - THE PROGRAM  ASSUMES  THAT  IT  IS  RUNNING
                UNATTENDED.  







                             FIV-PARAMETERS-PAGE 8


                               HASSLE STAT PACK
                                  F IV - STAT


                THE  FOLLOWING  ARE  THE ACTIONS TAKEN BY THE SETTINGS
                OF THE ATT SWITCH:  

                TRUE                    FALSE
                KEYBOARD INPUT ALLOWED  NOT ALLOWED     NOT UTILIZED
                                                        AT PRESENT
                ERRORS TO 4             TRACE AND ERRORS TO 3
                PROMPTS 'ANALYSIS NO'   NO PROMPTING GIVEN


        LIST AND ATT ARE FALSE AT PROGRAM START.         

        LDEVE   IS THE DEVICE FOR ERROR TEXTS. IT IS  DEFAULTED  TO  3
                WHEN ATT IS FALSE AND 4 WHEN ATT IS TRUE.  

        LDEVEL   IS  THE  DEVICE  FOR  TRACE. IT IS  DEFAULTED TO 4 AT
                PROGRAM START.   

        NOPAGE  THE PROGRAM NORMALLY GIVES A NEW  PAGE  FOR  EACH  NEW
                DATA  REQUEST  (ANALYSIS  0).  IF  A  LARGE  NUMBER OF
                GROUPS IS TO BE REPEATED WITH  E.G.  ONLY  MEANS  THIS
                TAKES  A LOT OF PAPER. SETTING NOPAGE TO TRUE INHIBITS
                THE PAGINATION.  

        THESE 5 BITS OF INFORMATION CAN BE CHANGED BY THE 49 CARD.  

        **NOTE**THE 49 CARD IS READ AS AN ANALYSIS  REQUEST  AND  MUST
        THEREFORE  NOT COME BEFORE THE INITIAL REQUESTS FOR GROUPS AND
        COLUMNS. THE FORMAT IS:  


        49,ATT,LIST,LDEVE,LDEVL,NOPAGE 
        ATT,LIST AND NOPAGE ARE SET TO TRUE IF THEY ARE EQUAL TO 1  ON
        THE  49  CARD  THEY  ARE  SET  TO  FALSE.  THE  49 CARD CAN BE
        REPEATED AT ANY TIME.    


        LDEVE AND LDEVL CAN BE SET TO 3,4 OR 9. IF SET  TO  ANY  OTHER
        NUMBER  THEY TAKE THEIR DEFAULT VALUES. IF ANY BIT IS NON-ZERO
        BUT A DEVICE IS ZERO IT IS NOT CHANGED I.E.  
        49,1    DOES NOT CHANGE LDEVE OR LDEVEL          

        BUT IF THE WHOLE CARD IS ZERO:  
        49,    THEN THE DEVICES  ARE  SET  TO  THEIR  DEFAULT  VALUES.













                             FIV-PARAMETERS-PAGE 9


                               HASSLE STAT PACK
                                  F IV - STAT


E)RUNNING STAT 


STAT IS AVAILABLE AS LOADER IMAGE AND RUN BY:  
.EX STAT 
*INDATAFILE/6 
*PARAMETER FILE/5 
*LPT:/3         OPTIONAL 
$  

SOME   TIMING   CONSIDERATIONS:  STAT  IS  RUNABLE  IN  ANY   12K  OS8
CONFIGURATION BUT IT IS SLOW IF NO FPP AND DISK IS  AVAILABLE. DUE  TO
THE  HEAVY OVERLAYING IT IS VERY SLOW IN A DECTAPE SYSTEM. IF POSSIBLE
TRY TO REARRANGE THE MODULES FOR FASTER CALLING IF  YOU  HAVE  SUCH  A
SYSTEM (I.E. UTILIZE MORE THAN 12K OF CORE).  



F)CAPACITIES 
IN 12 K CORE MAXXL=1000.
COLUMNS IN THE FILE MAX 10.
TOTAL DATA IN CORE (COLUMNS * ROWS) < MAXXL.


FOR NON PARAMETRIC ANALYSES:


GIVEN THAT
NK=NO OF COLUMNS IN CORE
NR=NO OF ROWS IN CORE

SO MUST FOR ROUTINE     THIS PRODUCT BE LESS THAN MAXXL
4                       (NK+2)*NR
13                      (NK+5)*NR
14                      (NK+4)*NR
15     MISSING.AND.15,1 (NK+2*NK)*NR
        ELSE            (NK+NK)*NR
16                      (NK+5)*NR
17      MISSING         (NK+2*NK)*NR
        NO MISSING      (NK+NK)*NR
18                      (NK+4)*NR
19                      (NK+1)*NR
20  21                  (NK+2)*NR














                            F IV - RUNNING-PAGE 10


                               HASSLE STAT PACK
                                  F IV - STAT



G)ERROR HANDLING

ERROR CONDITION                         HANDLED         MESSAGE

A) ON READING INPUT DATA PARAMETERS:
COLUMN OR GROUP REQUEST OUT OF RANGE    JOB ABORTED$    YES
TOTAL DATA>MAXXL                        JOB ABORTED$    YES

B)ON READING THE ANALYSIS REQUESTS:
NON-EXISTING ANALYSIS REQUESTED         IGNORED         YES
ANALYSIS REQUIRES COL NR THESE ARE WRONG
                <TOO LARGE OR .LE.0>    IGNORED         #
ANALYSES REQUIRES A 3RD ARGUMENT 
BUT THIS IS MISSING                     IGNORED         #

C)IN SPECIFIC ANALYSES:
        BESIDE THESE THERE ARE SOME SELFEXPLANATORY MESSAGES.
NR
-1      ARGUMENT OUT OF RANGE           IGNORED         #
        BAD FORMAT ON 0                 FRTS ABORTS PROGRAM
1       ARG OUT OF RANGE                IGNORED         #
2       CODE WRONG                      IGNORED         YES
2       TOO FEW DATA                    IGNORED         YES
3       TOO FEW DATA                    IGNORED         YES
4       BAD P VALUE                     SKIP 4          #
7           "                           SKIP 7          #
3,6,8   EMPTY COL                       IGNORED         YES
13      N<6                             SKIPPED         #
15      NK<3                            SKIPPED         #
16      N<4                             SKIPPED         #
17      NK<3                            SKIPPED         #
18      N<3                             SKIPPED         #

22      ON BAD ARGUMENTS(<0)TO SQUARE ROOTS OR LOGS
        THE VALUE IS SET TO 0

MANY ROUTINES ON EXCEEDING CAPACITY     IGNORED         #


#ALL THESE ARE SIGNALLED BY THE TEXT:  
SPECIFICATION ERROR IN STAT## WHERE ## IS THE MODULE NUMBER.    















                              FIV - ERROR-PAGE 11


                               HASSLE STAT PACK
                                  F IV - STAT


ALL THESE TEXTS ARE SENT TO DEVICE LDEVE.   

IN  SOME  PLACES  A  DIVIDE  BY  ZERO MIGHT BE PRODUCED WHEN IT IS NOT
GUARDED FOR.  

NOTE THAT LINE NUMBERS ARE IRRELEVANT.   

$IF THE JOB IS ABORTED THE PROGRAM SEARCHES THE PARAMETER FILE  FOR  A
50  CARD  OR  A  0  CARD SO AS TO <IF POSSIBLE> CONTINUE WITH THE NEXT
INDATA REQUEST.    



H)LOADING 


THE LOADING INFORMATION IS IN THE FILE  STATB.BI  FOR  USE  BY  BATCH.
THE  FILE  STATC.BI  IS  A  BATCH  FILE  FOR  COMPILING  THE NECESSARY
ROUTINES (NOTE: THIS FILE IS OS/8 V 3 FORMAT).  

NOTE THAT ALL MODULES ARE COMPILED WITHOUT LINE NUMBERS.  THE  PROGRAM
REQUIRES  12K.   IF  THE  PROGRAM  IS LOADED WITHOUT OVERLAYS IT TAKES
ABOUT 35 K.  IN A 16K DECTAPE SYSTEM YOU ARE ADVICED TO MOVE  SOME  OF
THE  ROUTINES  NOW  AT  LEVEL  2  TO  LEVEL MAIN.  THIS SAVES A LOT OF
OVERLAYING CORE.  

IF YOU WANT TO REMOVE ROUTINES FROM THE PROGRAM <IF  YOU  FEEL  IT  IS
UNNECESSARILY  LARGE>  STUDY  THE  TRANSFER  TABLE  IN  SORT.  YOU CAN
INHIBIT CALLS TO THE VARIOUS MODULES BY CHANGING  THE  VALUES  IN  THE
DATA  LIST  IN  SORT.  NOTE THAT THE OVERLAYS ARE VERY COMPACT PACKED.
TO INCREASE THE PACKAGE A GOOD DEAL OF THOUGHT WOULD BE REQUIRED.  



I)TESTDATA 


TESTDATA TO THE  PROGRAM IS IN STATDA.DA(/6) 
USING STATST.DA AS UNIT 5 WILL TEST ALL ROUTINES IN THE PACKAGE.  



J)REFERENCES:  


BESIDES THOSE MENTIONED IN  THE  GENERAL  INTRODUCTION  THE  FOLLOWING
BOOKS HAVE BEEN REFERRED TO IN THIS PACKAGE:  


D.COLQUHOUN (C)                 LECTURES ON BIOSTATISTICS 1971 

M ABRAMOWITZ & I STEGUN         HANDBOOK OF 
            (A)                 MATHEMATICAL FUNCTIONS 1964 




                              FIV-LOADING-PAGE 12


                               HASSLE STAT PACK
                                  F IV - STAT


K)SUBROUTINES 


THE  STATISTICAL PACKAGE  USES  A NUMBER  OF  SUBROUTINES THAT  ARE OF
A MORE GENERAL NATURE AND AS SUCH  ARE  NOT  CONSIDERED  PART  OF  THE
PACKAGE.  THESE ARE:  


1)  A NUMBER OF ROUTINES FROM THE SSP LIBRARY (SCIENTIFIC SUBROUTINES)
SEE THE DEC-10 LIBRARY FOR FURTHER INFORMATION. THESE ROUTINES ARE:  

UTEST   (PERFORMS MANN WHITNEY UTEST) 
MPAIR   (PERFORMS WILCOXONS MATCHED PAIRS RANK SIGN TEST) 
SRANK   (PERFORMS SPEARMAN RANK) 
TWOAV   (PERFORMS FRIEDMAN ANALYSIS) 
AND THE GENERAL ROUTINES TIE, NDTR AND RANK.  


2)SEVERAL GENERALIZED SUBROUTINES PROGRAMMED FROM VARIOUS SOURCES:  

NUMINT          NUMERICAL INTEGRATION   (D&R) 
PVALUE          PROBABILITY FUNCTION    (ADAPTED FROM DEC10 LIBRARY) 
FINV,TINV,ZINV  PROBABILITY FUNCTIONS   (A) 


THE OTHER SUBROUTINES ARE  SPECIFIC  TO  THE  PACKAGE.   ALL  ROUTINES
SPECIFIC TO THE PACKAGE HAVE NAMES IN THE FORM STAT??.FT.    






























                            FIV-SUBROUTINES-PAGE 13