1
MULTIVARIATE ANALYSIS
Multivariate analysis Elaborasi contingency tables split correlation analysis high-order partial correlation path analysis
Regressi/ Prediksi multiple regression
Differensiasi discriminant analysis manova
Eksplorasi/ identifikasi factor analysis cluster analysis
Elaborasi98/10/28/08
2
Elaborasi • Elaborasi:
vr.
Memasukkan (mengontrol/membuat konstan) variabel ketiga dalam analisis hubungan antara independent vr. dan dependent guna merinci hubungan tsb.
•
Variabel ketiga:
test factor/variable, intervening vr., antecedent vr., specifying vr., dsb.
(tergantung Kerangka Teori)
•
(merubah konstanta)
Pengontrolan variabel:
membuat konstan suatu vr. suatu variabel menjadi
• Beberapa teknik elaborasi: 1. contingency tables Elaborasi98/10/28/08
3
2. split/differential analysis 3. high-order partial correlation
Kemungkinan Hasil Suatu Elaborasi: Konstan
Replication Melemah
Explanation 3rd vr = antecedent
vr.
Interpretation
3rd vr = intervening vr.
Terbelah
Specification 3rd vr = specifying vr.
Menguat
Antecent Elaborasi98/10/28/08
3rd vr = suppressor vr.
Independent Intervening
4
Dependent variable(s) variable(s)
variable(s)
extraneous variable(s)
Elaborasi98/10/28/08
variable(s)
Specifying variable(s)
5
Elaborasi98/10/28/08
6
Elaborasi 1
CONTINGENCY TABLES Tabel Silang SES dan Konservatisme ( Zero-order Table )
n = 500 Pria & Wanita
SES Konservatisme Tinggi Rendah
Rendah 33% 67% 100%
Tinggi 78% 22% 100%
Tabel Silang SES dan Konservatisme Dalam Kelompok Pria ( First-order Table )
n = 300 Pria
SES Konservatisme Tinggi Rendah
Rendah 48% 52% 100%
Tinggi 49% 51% 100%
Tabel Silang SES dan Konservatisme Dalam Kelompok Wanita ( First-order Table )
n = 200 Wanita
SES Konservatisme Tinggi Rendah
Elaborasi98/10/28/08
Rendah 13% 87% 100%
Tinggi 81% 19% 100%
7
Elaborasi 1
Contingency Tables Zero-order Table
LMIDSTAT minat statistik by LMPS minat mps
LMPS
Count ” Row Pct ”tidak be berminat Col Pct ”rminat Row ” 1” 2” Total LMIDSTAT ““““““““•““““““““•““““““““› 1 ” 18 ” 50 ” 68 tidak berminat ” 26.5 ” 73.5 ” 60.7 ” 75.0 ” 56.8 ” š““““““““•““““““““› 2 ” 6 ” 38 ” 44 berminat ” 13.6 ” 86.4 ” 39.3 ” 25.0 ” 43.2 ” –““““““““ン ““““““““˜ Column 24 88 112 Total 21.4 78.6 100.0 Chi-Square ---------------------------Continuity Correction Likelihood Ratio Minimum Expected Frequency -
Elaborasi98/10/28/08
Value ----------1.90682 2.73749 9.429
DF
Significance ---1 1
.16732 .09802
8 08 May 96 SPSS for MS WINDOWS Release 6.0
First-order Tables ( controling for sex ) SEX:
1
LMIDSTAT
pria minat statistik
by
LMPS
LMPS
minat mps
Page 1 of 1
Count ” Row Pct ”tidak be berminat Col Pct ”rminat Row ” 1” 2” Total LMIDSTAT ““““““““•““““““““•““““““““› 1 ” 8 ” 24 ” 32 tidak berminat ” 25.0 ” 75.0 ” 55.2 ” 100.0 ” 48.0 ” š““““““““•““““““““› 2 ” ” 26 ” 26 berminat ” ” 100.0 ” 44.8 ” ” 52.0 ” –““““““““ン ““““““““˜ Column 8 50 58 Total 13.8 86.2 100.0 Chi-Square -------------------Continuity Correction Likelihood Ratio Fisher's Exact Test:
Value ----------5.58407 10.54857 One-Tail Two-Tail Minimum Expected Frequency 3.586 Cells with Expected Frequency < 5 2 OF
8 May 96 SPSS for MS WINDOWS Release 6.0
Elaborasi98/10/28/08
DF ---1 1
4 ( 50.0%)
Significance --------.01812 .00116 .00549 .00630
9
First-order Tables ( controling for sex ) SEX:
2
wanita
LMIDSTAT
minat statistik
by
LMPS
LMPS
minat mps
Page 1 of 1
Count ” Row Pct ”tidak be berminat Col Pct ”rminat Row ” 1” 2” Total LMIDSTAT ““““““““•““““““““•““““““““› 1 ” 10 ” 26 ” 36 tidak berminat ” 27.8 ” 72.2 ” 66.7 ” 62.5 ” 68.4 ” š““““““““•““““““““› 2 ” 6 ” 12 ” 18 berminat ” 33.3 ” 66.7 ” 33.3 ” 37.5 ” 31.6 ” –““““““““ン ““““““““˜ Column 16 38 54 Total 29.6 70.4 100.0 Chi-Square ------------------Pearson Continuity Correction Likelihood Ratio Mantel-Haenszel test for linear association Minimum Expected Frequency -
Value ----------.17763 .01110 .17574 .17434
Elaborasi98/10/28/08
5.333
DF 1 1 1 1
----
Significance --------.67342 .91609 .67506 .67628
10 08 May 96 SPSS for MS WINDOWS Release 6.0
Elaborasi 2 SPLIT CORRELATION ANALYSIS Zero-order Correlation MIDSTAT
MPS
MIDSTAT
1.0000 ( 112) P= .
.5928 ( 112) P= .000
MPS
.5928 ( 112) P= .000
1.0000 ( 112) P= .
First-order Correlation (controlling for sex) SEX:
1
pria MIDSTAT
SEX:
2
MPS
MIDSTAT
1.0000 ( 58) P= .
.8099 ( 58) P= .000
MPS
.8099 ( 58) P= .000
1.0000 ( 58) P= .
wanita
MIDSTAT
MPS
MIDSTAT
1.0000 ( 54) P= .
.2321 ( 54) P= .091
MPS
.2321 ( 54) P= .091
1.0000 ( 54) P= .
Elaborasi98/10/28/08
11
Contoh HIGH-ORDER P A R T I A L ANALYSIS
CORRELATION
Zero-order Correlation Coefficient MIDSTAT
MPS
MIDSTAT
1.0000 ( 112) P= .
.5928 ( 112) P= .000
MPS
.5928 ( 112) P= .000
1.0000 ( 112) P= .
First-order Partial Correlation Coefficient Controlling for..
LOGIKA MPS
MIDSTAT
MPS
1.0000 ( 0) P= .
.1687 ( 109) P= .077
MIDSTAT
.1687 ( 109) P= .077
1.0000 ( 0) P= .
Elaborasi98/10/28/08
12
- -
Matrix
Correlation Coefficients
- -
MIDSTAT
MPS
LOGIKA
MIDSTAT
1.0000 ( 112) P= .
.5928 ( 112) P= .000
.8054 ( 112) P= .000
MPS
.5928 ( 112) P= .000
1.0000 ( 112) P= .
.6407 ( 112) P= .000
LOGIKA
.8054 ( 112) P= .000
.6407 ( 112) P= .000
1.0000 ( 112) P= .
(Coefficient / (Cases) / 2-tailed Significance) " . " is printed if a coefficient cannot be computed
Elaborasi98/10/28/08
13
Elaborasi 3
HIGH-ORDER PARTIAL CORRELATION ANALYSIS • Zero-order Coefficient
Partial
N∑ xy
-
Correlation
(∑x) (∑y)
rxy = [ N∑x2 - (∑x)2 ] [ N∑y2 - (∑y)2 ]
•
First-order Coefficient
Partial Correlation (Controling for A)
rXY - (rXA)(rYA) rXY/A = (1- rXA2) (1- rYA2)
•
Second-order Partial Correlation Coefficient (Controling for A and B) r
XY.A
-
(rXB.A) (rYB.A)
rXY/AB = (1- rXB.A2) (1- rYB.A2) Elaborasi98/10/28/08
14
• Third-order Partial Coefficient etc. etc.
Correlation
Uji Signifikansi Statistik Perbedaan Dua Koefisien Korelasi (Marascuilo and Serlin, 1988; p.646)
H0 : ρ = 0
t=
r
N-2-V 1 - r2
Tolak H0 bila t > tN-2-V; α
(two-tailed test)
H0 : ρ = ρ 0 dengan transformasi ke nilai Z
Z = Zr - Z 0
1/ √N-3-V tolak H0 bila Z ≥ 1.96 (bila ∝ = 0.05)
H0 : ρ 1 = ρ 2 dengan transformasi ke nilai Z
Z r - Z0 Z
Elaborasi98/10/28/08
=
1
+
1
15
N1-3
N2-3
tolak H0 bila Z ≥ 1.96 (bila ∝ = 0.05)
Simple Regression • The use of sample statistics to predict population parameters • The use of past and current data to predict future results linear simple regression equation
y| = a +
= the value of y when x
y-intercept (constant) =0
a =
bx
y- b(x)
slope =the number of unit changed in y for every increase (or decrease of one unit in x
N∑ xy b =
Elaborasi98/10/28/08
-
(∑x) (∑y) = N ∑x2 - (∑x)2
16
linear simple regression equation
y| = a +
bx
Trend Analysis The use of past and current data to predict future results Tahun saham 1989 1990 1991 1992 1993 1994 1995 1996
Keuntungan per satuan 0.96 1.03 1.23 1.15 1.10 1.40 2.10 ?
2000
Tahu n 1989 1990 1991 1992 1993 1994
?
x
y
x2
y2
xy
1 2 3 4 5 6
.96 1.03 1.23 1.15 1.10 1.40
1 4 9 16 25 36
.92 1.06 1.51 1.32 1.21 1.96
.96 2.06 3.69 4.60 5.50 8.40
Elaborasi98/10/28/08
17
1995
7 ∑x = 28
2.10 ∑y = 8.97
49 ∑x2 = 140
4.41 ∑y2 = 12.39
14.70 ∑xy = 39.91
Trend Analysis Tahu n 1989 1990 1991 1992 1993 1994 1995 N= 7
x
y
x2
y2
xy
1 2 3 4 5 6 7 ∑x = 28
.96 1.03 1.23 1.15 1.10 1.40 2.10 ∑y = 8.97
1 4 9 16 25 36 49 ∑x2 = 140
.92 1.06 1.51 1.32 1.21 1.96 4.41 ∑y2 = 12.39
.96 2.06 3.69 4.60 5.50 8.40 14.70 ∑xy = 39.91
N∑ xy
- (∑x) (∑y)
rxy =
= [ N∑x2 - (∑x)2 ] [ N∑y2 - (∑y)2 ] N∑ xy - (∑x) (∑y)
b =
= 0.14 N ∑x - (∑x) 2
Elaborasi98/10/28/08
2
0.81 reject H0
(r0.05;5= 0.75)
18
a = y- b(x)
= 0.721
y| = a + bx
•
)
Estimasi profit 1996 (tahun ke-8) = bx = 0.721 =
•
+ 0.14 ( 8
1.84
Estimasi profit 2001 (tahun ke-12) + 0.14 ( 12 ) = 2.41
Elaborasi98/10/28/08
y| = a +
=
0.721
19
13 May 96 SPSS for MS WINDOWS Release 6.0
Page 1
****REGRESSION **** Listwise Deletion of Missing Data Equation Number 1 Dependent Variable.. Block Number 1. Method: Enter YEAR
PROFIT
Variable(s) Entered on Step Number 1.. YEAR Multiple R R Square Adjusted R Square Standard Error
.80125 .64199 .57039 .25434
Analysis of Variance DF 1 5
Regression Residual F =
8.96623
Sum of Squares .58003 .32345 Signif F =
Mean Square .58003 .06469
.0303
---------------------- Variables in the Equation -------------Variable Beta YEAR .801245 (Constant)
B
SE B
95% Confdnce Intrvl B
.143929
.048066
.020372
.267485
.705714
.214960
.153151
1.258277
End Block Number
Elaborasi98/10/28/08
1
All requested variables entered.
20
Standard Error of Estimate
Sest
= Sy
1 - r2
∑ y2 Sy =
-
( y )2 = 0.36
N
Sest
= 0.36
1 - ( 0.81)2 = 0.21
Accuracy Limits / Confidence Interval Limits ( alpha= 0.05 confidence interval = 95%)
AL.95 = = =
y| [ ± 1.96 ) (Sest ) ] 1.84 [ ± 1.96 ) ( 0.21 ) ] 1.84 ± 0.41
95% confidence y| (1996 profit per satuan saham) = 1.42 ≤ y| ≤ 2.25
Elaborasi98/10/28/08
21 12 May 96 SPSS for MS WINDOWS Release 6.0
Page 1
M U LT I PLE R E G R E S S I O N (Method: Enter) Equation Number 1
Dependent Variable..
Block Number 1. Method: Enter LOGIKA METODE TOEFL TPA Variable(s) Entered 1.. STATSOS 2.. TPA 3.. TOEFL 4.. METODE 5.. LOGIKA
STATSOS
.76899 .59135 .57208 .25410
Analysis of Variance
F =
indeks prestasi kumulatif
on Step Number test statistik test potensi akademik predict TOEFL test metodologi riset test logika
Multiple R R Square Adjusted R Square Standard Error
DF 5 106
Regression Residual
IPK
Sum of Squares 9.90429 6.84428
30.67831
Signif F =
Mean Square 1.98086 .06457
.0000
---------------------- Variables in the Equation ----------------------Variable LOGIKA METODE TOEFL TPA STATSOS (Constant)
B .016074 9.01291E-04 4.23286E-04 .003154 -.034613 1.169435
SE B
95% Confdnce Intrvl B
.006239 .003704 .003307 -.005655 5.8488E-04 -7.36286E-04 5.1792E-04 .002127 .084556 -.202254 .179047 .814457
.028443 .007457 .001583 .004181 .133029 1.524414
Beta .289058 .029872 .074931 .547330 -.044034
----------- in -----------Variable LOGIKA METODE TOEFL TPA STATSOS (Constant)
T
Sig T
2.576 .273 .724 6.090 -.409 6.531
.0114 .7857 .4708 .0000 .6831 .0000
Equation Number 1 End Block Number
Elaborasi98/10/28/08
Dependent Variable.. 1
IPK
indeks prestasi kumulatif
All requested variables entered
22
12 May 96 SPSS for MS WINDOWS Release 6.0
Page 2
M U LT I PLE R E G R E S S I O N (Method: Stepwise) Equation Number 1 kumulatif
Dependent Variable..
Block Number 1. Method: Stepwise .1000 LOGIKA METODE TOEFL TPA
IPK
Criteria
indeks prestasi PIN
.0500
POUT
STATSOS
Variable(s) Entered on Step Number 1.. TPA test potensi akademik Multiple R R Square Adjusted R Square Standard Error
.72113 .52003 .51567 .27033
Analysis of Variance
DF 1 110
Regression Residual F =
119.18174
Sum of Squares 8.70979 8.03879 Signif F =
Mean Square 8.70979 .07308
.0000
---------------------- Variables in the Equation ----------------------Variable TPA (Constant)
B
SE B
.004155 1.117701
3.8064E-04 .158138
----------- in -----------Variable TPA (Constant)
T
Sig T
10.917 7.068
.0000 .0000
Elaborasi98/10/28/08
95% Confdnce Intrvl B .003401 .804309
.004910 1.431094
Beta .721132
23
12 May 96 SPSS for MS WINDOWS Release 6.0
* * * *
Page 3
M U L T I P L E
Equation Number 1 kumulatif
R E G R E S S I O N
Dependent Variable..
IPK
* * * *
indeks prestasi
Variable(s) Entered on Step Number 2.. LOGIKA test logika Multiple R R Square Adjusted R Square Standard Error
.76642 .58740 .57983 .25179
Analysis of Variance DF 2 109
Regression Residual F =
77.58984
Sum of Squares 9.83814 6.91043 Signif F =
Mean Square 4.91907 .06340
.0000
--------------------- Variables in the Equation ----------------------Variable LOGIKA TPA (Constant)
B
SE B
.015826 .003483 1.230336
.003751 3.8875E-04 .149691
95% Confdnce Intrvl B .008391 .002712 .933652
.023261 .004253 1.527019
----------- in -----------Variable LOGIKA TPA (Constant)
T
Sig T
4.219 8.959 8.219
.0001 .0000 .0000
------------- Variables not in the Equation ------------Variable METODE TOEFL STATSOS
Beta In
Partial
Min Toler
T
Sig T
.061428 .061758 .080414 .086330 -.020683 -.019048
.417039 .475540 .335437
.643 .901 -.198
.5216 .3698 .8434
Elaborasi98/10/28/08
Beta .284609 .604377
24 End Block Number
1
PIN =
.050 Limits reached.
12 May 96 SPSS for MS WINDOWS Release 6.0
Page 5
Multiple r
r1y2 + r2y2 - 2(r1y r2y r12) r 12.y = 1 - r122
y = midstat 1 = MPS
2 = Logika
r1y = 0.5928
r2y = 0.6407
r12 = 0.8054
(0.5928)2 + (0.6407)2 - 2[(0.5928)(0.6407)(0.8054) r 12.y = 1 - (0.8054)2
Elaborasi98/10/28/08
25
Multiple Regression y| = a + b1x1 + b2x2 + . . . Sy
r1y - r2y r12
S1
1 - r122
Sy
r2y - r1yr12
S2
1 - r122
b1 =
b2 =
a =
y - b1x1 - b2x2
Standard Error of Multiple Estimate
Sm.est =
1 - r 12.y2
AL.95 =
y| ± 1.96 (Sm.est)
Elaborasi98/10/28/08
26
Elaborasi98/10/28/08
27
Elaborasi98/10/28/08