Multiple Classification Analysis using SPSS Widyo Pura Buana
Widyo Pura Pura Buana Buana - MCA
TEKNIK VARIABEL TERPENGARUH / VARIABEL DEPENDEN DEPEN DEN VA VARIAB RIABEL EL (Y)
ANALISIS DATA VARIABEL VARIAB EL PENGARUH / INDEPENDEN INDEPENDEN VARIABEL VARIABEL (X) NOMINAL
Dikotomi NOMINAL
Dikotomi
1. 2. 3. 4.
Difference of proportion test Chi-square Fish Fishe ers rs ex exact act tes testt Phi coefficie cient
Poli Polittomi omi
1. 1. 2.
ChiChi-sq squa uarre Kendalls VCT
ORDINAL
1. 2.
Man-Whitney Smirnov-Kolmogoronov
INTERVAL
1. 2.
Analysis of variance Diff Differen erence ce of of mean meanss test test (Sch (Schef effe fe test) Sign test M-test U-test Cros Crosss-cla class ssif ific icat atio ion n anal analys ysis is
3. 4. 5. 6.
Widyo Pura Pura Buana Buana - MCA
Politomi
1. 2.
Chi-square Kendalls VCT
1.
Anal Analys ysis is of of var varia iance nce wit with h inte interc rcla lass correlation Dumm Dummy y varia ariabl ble es mult multip iple le regression Mult Multip iple le clas classi sific ficat atio ion n anal analys ysis is Cros Crosss-cla class ssif ific icat atio ion n anal analys ysis is
2. 3. 4.
TEKNIK VARIABEL TERPENGARUH / VARIABEL DEPENDEN DEPEN DEN VA VARIAB RIABEL EL (Y)
ANALISIS DATA VARIABEL VARIAB EL PENGARUH / INDEPENDEN INDEPENDEN VARIABEL VARIABEL (X) NOMINAL
Dikotomi NOMINAL
Dikotomi
1. 2. 3. 4.
Difference of proportion test Chi-square Fish Fishe ers rs ex exact act tes testt Phi coefficie cient
Poli Polittomi omi
1. 1. 2.
ChiChi-sq squa uarre Kendalls VCT
ORDINAL
1. 2.
Man-Whitney Smirnov-Kolmogoronov
INTERVAL
1. 2.
Analysis of variance Diff Differen erence ce of of mean meanss test test (Sch (Schef effe fe test) Sign test M-test U-test Cros Crosss-cla class ssif ific icat atio ion n anal analys ysis is
3. 4. 5. 6.
Widyo Pura Pura Buana Buana - MCA
Politomi
1. 2.
Chi-square Kendalls VCT
1.
Anal Analys ysis is of of var varia iance nce wit with h inte interc rcla lass correlation Dumm Dummy y varia ariabl ble es mult multip iple le regression Mult Multip iple le clas classi sific ficat atio ion n anal analys ysis is Cros Crosss-cla class ssif ific icat atio ion n anal analys ysis is
2. 3. 4.
TEKNIK VARIABEL TERPENGARUH / DEPENDEN VARIABEL (Y) NOMINAL
Dikotomi
ANALISIS DATA
VARIABEL PENGARUH / INDEPENDEN VARIABEL VARIABEL (X) ORDINAL
1. Kruskall-Wallis 2. Frie Friedma dman nss 2 way way analy analysi siss of var varia ianc nce e
Politomi ORDINAL
1. Rank-order correlation 2. Kendalls tau 3. Gamma 4. Coef Coeffi fici cien entt of conc concor orda danc nce e
INTERVAL
Ubah var ordinal ja jadi var nominal & pakai analysis of variance, DVM R, MCA atau Ubah var interval Jadi var ordinal & pakai statistik statistik non-parametrik
Widyo Pura Pura Buana Buana - MCA
TEKNIK VARIABEL TERPENGARUH / DEPENDEN VARIABEL (Y) NOMINAL
Dikotomi
ANALISIS DATA
VARIABEL PENGARUH / INDEPENDEN VARIABEL (X) INTERVAL
1. Logistic multiple regression 2. Discriminant analysis
Politomi ORDINAL
Ubah var ordinal jadi var nominal & pakai logistic multiple regression & discriminant analysis atau Ubah var interval jadi var ordinal & pakai statistik non-parametrik
INTERVAL
1. 2. 3. 4.
Correlation atau regression Multiple correlation atau multiple regression Path analysis Partial regression
Widyo Pura Buana - MCA
Multiple Regression and Multiple Classification Analysis Introduction This chapter examines a model of multivariate analysis, involving simultaneous consideration of several independent (predictor or explanatory) variables and one dependent variable, where the objectives of analysis are: (i) To know how well all the independent variables together explain variation in the dependent variable. (ii) To know how well each independent variable is related to the dependent variable, either considering or ignoring the effects of other independent variables. Widyo Pura Buana - MCA
Multiple Regression and Multiple Classification Analysis The
following data analysis situations can be visualized, depending upon the measurement properties of the dependent and independent variables. Dependent variable
I ndependent
variables Statistical techniques
One
Several
Interval scale
Interval scale
Multiple Regression
Interval scale
Nominal
Multiple Classification Analysis
Dichotomous,
Nominal
Multiple Classification Analysis
Polytomous Widyo Pura Buana - MCA
Multiple Classification Analysis (MCA) Multiple Classification Analysis (MCA) is a technique for examining the interrelationship between several predictor variables and one dependent variable in the context of an additive model. Unlike simpler forms of other multivariate methods, MCA can handle predictors with no better than nominal measurements and interrelationships of any form among the predictor variables or between a predictor and dependent variable. It is however essential that the dependent variable should be either an interval-scale variable without extreme skewness or a dichotomous variable with frequencies which are not extremely unequal. Widyo Pura Buana - MCA
Model MCA Y ij...n=
Y + a i +b j+ . . . .+e ij..n
where Y ij...n = The score on the dependent variable of individual n who falls in category i of predictor A, category j of predictor B, etc
Y
= Grand mean of the dependent variable.
a i = The effect of the membership in the i th category of predictor A. b j = The effect of the membership in the j th category of predictor B. e ij..n= Error term for this individual.
Widyo Pura Buana - MCA
Model MCA Y ij ...n ! Grand Mean Row Effect Column Effect Residual
Grand Mean
Y ij ...n !
Row Effect
Column Effect Residual
Widyo Pura Buana - MCA
Performance by Task Difficulty and Arousal
Arousal (Column) Low Task Difficulty (Row)
Easy
Difficult
Column Mean
Medium High
3
2
9
1
5
9
1
9
13
6
7
6
4
7
8
0
3
0
2
8
0
0
3
0
0
3
5
3
3
0
2
5
5
Widyo Pura Buana - MCA
Row Mean
6
2
4
Grand Mean
2
3
SS Total ! §§ ( yij Y ) 2 i !1 j !1
! (3 - 4) 2 ... (0 4) 2 ! 360 2
SS o ! §
2 ( y Y ) i i.
i !1 2
2
! 15.(6 4) 15.( 2 4) ! 30 30 ! 60 3
SS olu
n
!§
j ( y. j Y )
2
j !1 2
2
! 10.( 2 4) 10.(5 4) 10.(5 4) ! 40 10 10 ! 60 Widyo Pura Buana - MCA
2
! SS o SS C olu
n
SS Model ! SS o SS C olu
n
SS C o
bined
SS R e sidual ! SS otal SS Model Widyo Pura Buana - MCA
df Total ! N 1 ! 30 1 ! 29 df o ! # of ro s (levels) 1 ! 2 1 ! 1 df C olu
n
! # of colu ns (levels) 1 ! 3 1 ! 2
df Combined ! df Row df Column ! 1 2 ! 3
df Model ! df Row df Column ! 1 2 ! 3 df Re sidual ! df Total df Model ! 29 3 ! 26 Widyo Pura Buana - MCA
Eta
(L)
Etarow ! L row !
Etacolumn ! L column ! Widyo Pura Buana - MCA
SS Row SS Total SS Column SS Total
Goodness of Fit
R Squared !
!
SS Model SS Total
Squared ! Widyo Pura Buana - MCA
SS
odel
SS T otal
Syntax SPSS MCA *MCA model with categorical predictors:. ANOVA Performance by Difficulty (1,2) Arousal (1,3) /MAXORDERS=NONE /METHOD=EXPERIMENTAL /STATISTICS=MCA.
Widyo Pura Buana - MCA
Struktur Data MCA dengan SPSS
Widyo Pura Buana - MCA
ANOVAa Tingkat Kesulitan Pekerjaan
dan Gairah Kerja berpengaruh terhadap Performance Kerja (baik secara overall atau individual)
Performance
Experimental Method Sum of Squares 180.000
3
Mean Square 60.000
120.000
1
60.000
Model
8.667
Sig. .000
120.000
17.333
.000
2
30.000
4.333
.024
180.000
3
60.000
8.667
.000
Residual
180.000
26
6.923
Total
360.000
29
12.414
Main Effects
(Combined) Task Difficulty Arousal
df
a. Performance by Task Difficulty, Arousal
Widyo Pura Buana - MCA
F
Significant
MCAa Predicted Mean Adjusted
N
Performance
Task Difficulty
Arousal
Deviation
Unadjusted
Adjusted
for Unadjusted Factors 6.00 2.000
for Factors 2.000
Easy
15
6.00
Difficult
15
2.00
2.00
-2.000
-2.000
Low
10
2.00
2.00
-2.000
-2.000
Medium
10
5.00
5.00
1.000
1.000
High
10
5.00
5.00
1.000
1.000
a. Performance by Task Difficulty, Arousal
Performance Deviation Mean Row
Column
Task
Difficulty Arousal
Easy Difficult Low Medium High Grand
Mean
6
±4 =2±4 =2±4 =5±4 =5±4
2 =6
2
-2
2
-2
5
1
5
1
4
Widyo Pura Buana - MCA
Row(i)-Grand
Mean
Column(j)-Grand Mean
Factor
Summarya Eta
Beta Adjusted for Factors
Task
Difficulty Performance (Row) Arousal (Column)
.577
.577
Formula =SQ RT( SSRow/ SSTotal ) =SQRT(120/360)
.408
.408
Widyo Pura Buana - MCA
=SQ RT( SSColumn/ SSTotal ) =SQRT(60/360)
Model Goodness of Fit R
Performance by Task Difficulty, Arousal
R
Squared
.707
.500
=SQ RT(R-Squared)
= SSModel/SSTotal
Widyo Pura Buana - MCA
Multiple Classification Analysis with Interaction
Widyo Pura Buana - MCA
Syntax SPSS MCA *MCA model with categorical predictors, interaction:. ANOVA Performance by Difficulty (1,2) Arousal (1,3) /MAXORDERS=ALL /METHOD=EXPERIMENTAL /STATISTICS=MCA.
Widyo Pura Buana - MCA
ANOVAa
Performance
Main Effects
(Combined) Task Difficulty
2-Way
Interactions Model Residual Total
Sum of Squares 180.000
Experimental Method Mean Square df F 3 60.000 12.000
Sig. .000
120.000
1
120.000
24.000
.000
Arousal
60.000
2
30.000
6.000
Task
60.000
2
30.000
6.000
.008 .008
240.000
5
48.000
9.600
.000
120.000
24
5.000
360.000
29
12.414
Difficulty * Arousal
Widyo Pura Buana - MCA
Graphical display of interactions Two
ways to display previous results iffic lt
A
¨
¦
10.00
§
©
iffic lt
10.00
§
8.00
8.00
r
6.00
r
2.00
2.00
¥
6.00
4.00
4.00
0.00
.00
s ¨
l
£
¤ ¡
Ar
¢
i
l
s ¨
s l
§
i
©
l
l
Widyo Pura Buana - MCA
iffic lt §
©
MCA $ GLM Factorial Anova
Widyo Pura Buana - MCA
MCA $ GLM Factorial Anova MULTIPLE CLASSIFICATION ANALYSIS (MCA) Melissa A. Hardy & Chardie L. Baird
MULTIPLE CLASSIFICATION ANALYSIS (MCA) Also called factorial A NOVA, multiple classification analysis (MCA) is a QUANTITATIVE analysis procedure that allows the assessment of differences in subgroup means, which may have been adjusted for compositional differences in related factors and/or covariates and their effects. MCA produces the same overall results as MULTIPLE REGRESSION with DUMMY VARIABLES, although there are differences in the way the information is reported. For example, an MCA in SPSS produces an ANALYSIS OF VARIANCE with the appropriate F TESTS, decomposing the SUMS OF SQUARES explained by the model into the relative contributions of the factor of interest, the C OVARIATE(s), and any INTERACTIONS that are specified. These F tests assess the ratio of the sums of squares explained by the factor(s) and covariates (if specified) adjusted... Source : http://srmo.sagepub.com/view/the-sage-encyclopedia-of-social-science-research-methods/n597.xml
Widyo Pura Buana - MCA
Graphical display of interactions What are we looking for? Do the lines behave similarly (are parallel) or not? Does the effect of one factor depend on the level of the other factor?
No
interaction
The lines are parallel
Interaction Widyo Pura Buana - MCA
The
lines are not parallel
GLM Factorial ANOVA Statistical
Model:
Y ijk ! Q E i F j EF ij I k ( ij ) Statistical Hypothesis:
Treatment A. H 0 : E1 ! E 2 ! L ! E p Treatment B. H 0 : F1 ! F 2 ! L ! F q Interaction. H 0 : EF11 ! EF12 ! L ! EF pq The interaction null is that the cell means d o not differ significantly (fr om the grand mean) outside of the main effects present, i.e. that this residual effect is zer o Widyo Pura Buana - MCA
Widyo Pura Buana - MCA
Widyo Pura Buana - MCA
Widyo Pura Buana - MCA
Widyo Pura Buana - MCA