C BISE 10
C ..................................................................BISE 20
C BISE 30
C SUBROUTINE BISER BISE 40
C BISE 50
C PURPOSE BISE 60
C TO COMPUTE THE BISERIAL CORRELATION COEFFICIENT BETWEEN TWO BISE 70
C CONTINUOUS VARIABLES WHEN ONE OF THEM HAS BEEN ARTIFICIALLY BISE 80
C DICHOTOMIZED. BISE 90
C BISE 100
C USAGE BISE 110
C CALL BISER (N,A,B,HI,ANS,IER) BISE 120
C BISE 130
C DESCRIPTION OF PARAMETERS BISE 140
C N - NUMBER OF OBSERVATIONS BISE 150
C A - INPUT VECTOR OF LENGTH N CONTAINING THE CONTINUOUS BISE 160
C VARIABLE BISE 170
C B - INPUT VECTOR OF LENGTH N CONTAINING THE DICHOTOMIZED BISE 180
C VARIABLE BISE 190
C HI - INPUT - NUMERICAL CODE TO INDICATE THE HIGHER CATEGORYBISE 200
C OF THE DICHOTOMIZED VARIABLE. ANY VALUE IN VECTOR B BISE 210
C EQUAL TO OR GREATER THAN HI WILL BE CLASSIFIED INTO BISE 220
C THE HIGHER CATEGORY. BISE 230
C ANS - OUTPUT VECTOR OF LENGTH 8 CONTAINING THE FOLLOWING BISE 240
C ANS(1) - MEAN OF VARIABLE A BISE 250
C ANS(2) - STANDARD DEVIATION OF VARIABLE A BISE 260
C ANS(3) - PROPORTION OF THE CASES IN THE HIGHER BISE 270
C CATEGORY OF VARIABLE B BISE 280
C ANS(4) - PROPORTION OF THE CASES IN THE LOWER BISE 290
C CATEGORY OF VARIABLE B BISE 300
C ANS(5) - MEAN OF VARIABLE A FOR THOSE CASES FALLING BISE 310
C INTO THE HIGHER CATEGORY OF VARIABLE B BISE 320
C ANS(6) - MEAN OF VARIABLE A FOR THOSE CASES FALLING BISE 330
C INTO THE LOWER CATEGORY OF VARIABLE B BISE 340
C ANS(7) - BISERIAL CORRELATION COEFFICIENT BISE 350
C ANS(8) - STANDARD ERROR OF BISERIAL CORRELATION BISE 360
C COEFFICIENT BISE 370
C IER - 1, IF NO CASES ARE IN THE LOWER CATEGORY OF VARIABLE BISE 380
C B. BISE 390
C -1, IF ALL CASES ARE IN THE LOWER CATEGORY OF BISE 400
C VARIABLE B. BISE 410
C 0, OTHERWISE. BISE 420
C IF IER IS NON-ZERO, ANS(I)=10**75,I=5,...,8. BISE 430
C BISE 440
C REMARKS BISE 450
C THE VALUES OF THE DICHOTOMIZED VARIABLE, B, MUST BE IN BISE 460
C NUMERIC FORM. THEY CANNOR BE SPECIFIED BY MEANS OF BISE 470
C ALPHABETIC OR SPECIAL CHARACTERS. BISE 480
C BISE 490
C SUBROUTINES AND FUNCTION SUBPROGRAMS REQUIRED BISE 500
C NDTRI BISE 510
C BISE 520
C METHOD BISE 530
C REFER TO P. HORST, 'PSYCHOLOGICAL MEASUREMENT AND BISE 540
C PREDICTION', P.95-96 (WADSWORTH, 1966). BISE 550
C BISE 560
C ..................................................................BISE 570
C BISE 580
SUBROUTINE BISER (N,A,B,HI,ANS,IER) BISE 590
C BISE 600
DIMENSION A(1),B(1),ANS(1) BISE 610
C BISE 620
C COMPUTE MEAN AND STANDARD DEVIATION OF VARIABLE A BISE 630
C BISE 640
IER=0 BISE 650
SUM=0.0 BISE 660
SUM2=0.0 BISE 670
DO 10 I=1,N BISE 680
SUM=SUM+A(I) BISE 690
10 SUM2=SUM2+A(I)*A(I) BISE 700
FN=N BISE 710
ANS(1)=SUM/FN BISE 720
ANS(2)=(SUM2-ANS(1)*SUM)/(FN-1.0) BISE 730
ANS(2)= SQRT(ANS(2)) BISE 740
C BISE 750
C FIND PROPORTIONS OF CASES IN THE HIGHER AND LOWER CATEGORIES BISE 760
C BISE 770
P=0.0 BISE 780
SUM=0.0 BISE 790
SUM2=0.0 BISE 800
DO 30 I=1,N BISE 810
IF(B(I)-HI) 20, 25, 25 BISE 820
20 SUM2=SUM2+A(I) BISE 830
GO TO 30 BISE 840
25 P=P+1.0 BISE 850
SUM=SUM+A(I) BISE 860
30 CONTINUE BISE 870
ANS(4)=1.0 BISE 880
ANS(3)=0.0 BISE 890
Q=FN-P BISE 900
IF (P) 35,35,40 BISE 910
35 IER=-1 BISE 920
GO TO 50 BISE 930
40 ANS(5)=SUM/P BISE 940
IF (Q) 45,45,60 BISE 950
45 IER=1 BISE 960
ANS(4)=0.0 BISE 970
ANS(3)=1.0 BISE 980
50 DO 55 I=5,8 BISE 990
55 ANS(I)=1.7E38 BISE1000
GO TO 65 BISE1010
60 ANS(6)=SUM2/Q BISE1020
P=P/FN BISE1030
Q=1.0-P BISE1040
C BISE1050
C FIND ORDINATE OF THE NORMAL DISTRIBUTION CURVE AT THE POINT OF BISE1060
C DIVISION BETWEEN SEGMENTS CONTAINING P AND Q PROPORTIONS BISE1070
C BISE1080
CALL NDTRI (Q,X,Y,ER) BISE1090
C BISE1100
C COMPUTE THE BISERIAL COEFFICIENT OF CORRELATION BISE1110
C BISE1120
R=((ANS(5)-ANS(1))/ANS(2))*(P/Y) BISE1130
C BISE1140
C COMPUTE THE STANDARD ERROR OF R BISE1150
C BISE1160
ANS(8)=( SQRT(P*Q)/Y-R*R)/SQRT(FN) BISE1170
C BISE1180
C STORE RESULTS BISE1190
C BISE1200
ANS(3)=P BISE1210
ANS(4)=Q BISE1220
ANS(7)=R BISE1230
C BISE1240
65 RETURN BISE1250
END BISE1260