|
**********************************************************************
*** Reprodução do Quadro de uma ANOVA Unifactorial *** Valentim R. Alferes (University of Coimbra, Portugal) *** valferes@fpce.uc.pt
**********************************************************************
** MÉTODO 1: Ns Médias e Desvios-padrão.
*
Enter, row by row, N, Mean and Standard Deviation for the
*
groups or experimental treatments (in the example, you have
*
three treatments).
DATA LIST LIST
/n(F8.0) m(f8.2) sd(f8.2).
BEGIN DATA
6 10,50
2,30
7 11,34
3,60
8 15,17
4,10
END DATA.
COMPUTE a=$CASENUM.
LOOP id=1 TO
n.
XSAVE
OUTFILE=XOUT1.
END LOOP.
EXECUTE.
GET
FILE=XOUT1.
COMPUTE vd=m.
COMPUTE k=SQR((sd**2*(N-1))/2).
IF (id=1) vd=m+k.
IF (id=2) vd=m-k.
EXECUTE.
SUMMARIZE/TABLES=vd BY a/FORMAT=NOLIST TOTAL
/TITLE='Case Summaries'/CELLS=COUNT MEAN STDDEV VAR.
ONEWAY vd BY a.
**********************************************************************
** MÉTODO 2: Ns, Médias e Variâncias.
*
Enter, row by row, N, Mean and Variance for the
*
groups or experimental treatments (in the example, you have
*
three treatments).
DATA LIST LIST
/n(F8.0) m(f8.2) var(f8.2).
BEGIN DATA
6 10,50
5,29
7 11,34
12,96
8 15,17
16,81
END DATA.
COMPUTE a=$CASENUM.
LOOP id=1 TO
n.
XSAVE
OUTFILE=XOUT1.
END LOOP.
EXECUTE.
GET
FILE=XOUT1.
COMPUTE vd=m.
COMPUTE k=SQR((var*(N-1))/2).
IF (id=1) vd=m+k.
IF (id=2) vd=m-k.
EXECUTE.
SUMMARIZE/TABLES=vd BY a/FORMAT=NOLIST TOTAL
/TITLE='Case Summaries'/CELLS=COUNT MEAN STDDEV VAR.
ONEWAY vd BY a.
**********************************************************************
** MÉTODO 3: Ns, Médias e Média Quadrática
do Erro.
*
Enter, row by row, N and Mean for the groups or experimental
*
treatments (in the example, you have three treatments).
DATA LIST LIST
/n(F8.0) m(f8.0).
BEGIN DATA
6 10,50
7 11,34
8 15,17
END DATA.
* Enter Mean Square Error.
COMPUTE mse = 12.32667.
COMPUTE a=$CASENUM.
LOOP id=1 TO
n.
XSAVE
OUTFILE=XOUT1.
END LOOP.
EXECUTE.
GET
FILE=XOUT1.
COMPUTE vd=m.
COMPUTE k=SQR((mse*(N-1))/2).
IF (id=1) vd=m+k.
IF (id=2) vd=m-k.
EXECUTE.
SUMMARIZE/TABLES=vd BY a/FORMAT=NOLIST TOTAL
/TITLE='Case Summaries'/CELLS=COUNT MEAN.
ONEWAY vd BY a.
**********************************************************************
** MÉTODO 4: Médias, GL numerador, GL
denominador e Média Quadrática do Erro.
*
Enter, row by row, the Mean for the groups or experimental
*
treatments (in the example, you have three treatments).
DATA LIST LIST /m(f8.2). BEGIN DATA
10,50
11,34
15,17 END DATA. * Enter degrees of freedom for numerator (between groups df). COMPUTE dfnum=2. * Enter degrees of freedom for denominator (within groups * or MSError df). COMPUTE dfden=18. * Enter MSError.
COMPUTE mse=12.32667.
COMPUTE a=$CASENUM. COMPUTE n_tot=dfnum+dfden+1. COMPUTE n_treat=dfnum+1. COMPUTE nxi=RND(n_tot/n_treat). EXECUTE. CREATE cum=CSUM(nxi). SORT CASES BY cum(D). IF ($CASENUM=1) nxi=nxi-(cum-n_tot). EXECUTE. SORT CASES BY a(A).
LOOP id=1 TO
nxi.
XSAVE
OUTFILE=XOUT1.
END LOOP.
EXECUTE.
GET
FILE=XOUT1.
COMPUTE vd=m.
COMPUTE k=SQR((mse*(nxi-1))/2).
IF (id=1) vd=m+K.
IF (id=2) vd=m-K.
EXECUTE.
SUMMARIZE/TABLES=vd BY a/FORMAT=NOLIST TOTAL
/TITLE='Case Summaries'/CELLS=COUNT MEAN.
ONEWAY vd BY a.
**********************************************************************
** Note
** With Method 4 you don’t know the Ns per treatment from the original
** source. If they are equal, you will get the exact ANOVA Table.
Otherwise,
** as is the case in this example, you get only an approximation,
** because the syntax assumes that the Ns are equal in all treatments
** (if N_TOT/N_TREAT is an integer), or that they are equal in all
** treatments, except for the last one (if N_TOT/N_TREAT is not an
integer).
**
** The syntax bellow exemplifies the case for equal Ns per treatment.
**********************************************************************
** The following syntax does the ANOVA on raw data from Kirk
**(1995, p. 167, Table 5.2-I) and compares it with Methods 1 and 4.
** Analysis of raw data from Kirk (1995, p. 167, Table 5.2-I).
DATA LIST FREE /A(F8.0) VD(F8.0).
BEGIN DATA
1 4
1 6 1 3 1 3 1 1 1 3 1 2 1 2 2 4 2 5 2 4 2 3 2 2 2 3 2 4 2 3 3 5
3 6
3 5 3 4 3 3 3 4 3 3 3 4 4 3 4 5 4 6 4 5 4 6 4 7 4 8 4 10
END DATA.
ONEWAY
vd BY a/STATISTICS DESCRIPTIVES.
**
METHOD 1: Ns Means, and SDs.
*
Enter, row by row, N, Mean and Standard Deviation for the
* groups
or experimental treatments (in the example, you have
*
three treatments).
DATA LIST LIST
/n(F8.0) m(f8.2) sd(f8.6).
BEGIN DATA
8 3,00
1,511858
8 3,50
0,925820
8 4,25
1,035098
8 6,25
2,121320
END DATA.
COMPUTE a=$CASENUM.
LOOP id=1 TO
n.
XSAVE
OUTFILE=XOUT1.
END LOOP.
EXECUTE.
GET
FILE=XOUT1.
COMPUTE vd=m.
COMPUTE k=SQR((sd**2*(N-1))/2).
IF (id=1) vd=m+k.
IF (id=2) vd=m-k.
EXECUTE.
SUMMARIZE/TABLES=vd BY a/FORMAT=NOLIST TOTAL
/TITLE='Case Summaries'/CELLS=COUNT MEAN STDDEV VAR.
ONEWAY vd BY a.
**
METHOD 4: Means, DFnum, DFden, and MSError.
*
Enter, row by row, the Mean for the groups or experimental
*
treatments (in the example, you have three treatments).
DATA LIST LIST
/m(f8.2).
BEGIN DATA
3,00
3,50
4,25
6,25
END DATA.
* Enter
degrees of freedom for numerator (between groups df).
COMPUTE
dfnum=3.
* Enter
degrees of freedom for denominator (within groups
* or MSError
df).
COMPUTE
dfden=28.
* Enter
MSError.
COMPUTE mse=2.178571.
COMPUTE a=$CASENUM.
COMPUTE
n_tot=dfnum+dfden+1.
COMPUTE n_treat=dfnum+1.
COMPUTE
nxi=RND(n_tot/n_treat).
EXECUTE.
CREATE
cum=CSUM(nxi).
SORT CASES BY
cum(D).
IF
($CASENUM=1) nxi=nxi-(cum-n_tot).
EXECUTE.
SORT CASES BY a(A).
LOOP id=1 TO
nxi.
XSAVE
OUTFILE=XOUT1.
END LOOP.
EXECUTE.
GET
FILE=XOUT1.
COMPUTE vd=m.
COMPUTE k=SQR((mse*(nxi-1))/2).
IF (id=1) vd=m+K.
IF (id=2) vd=m-K.
EXECUTE.
SUMMARIZE/TABLES=vd BY a/FORMAT=NOLIST TOTAL
/TITLE='Case Summaries'/CELLS=COUNT MEAN.
ONEWAY vd BY a.
|