Professional Documents
Culture Documents
18-1
© 2007 Prentice Hall
Cha pter Ou tl ine
1) Overview
2) Basic Concept
3) Relation to Regression and ANOVA
4) Discriminant Analysis Model
5) Statistics Associated with Discriminant Analysis
6) Conducting Discriminant Analysis
i. Formulation
ii. Estimation
iii. Determination of Significance
iv. Interpretation
v. Validation
18-2
© 2007 Prentice Hall
Cha pter Ou tl ine
1) Multiple Discriminant Analysis
i. Formulation
ii. Estimation
iv. Interpretation
v. Validation
18-3
© 2007 Prentice Hall
Ch apter O utl ine
i. Estimation
v. An Illustrative Application
10) Summary
18-4
© 2007 Prentice Hall
Sim il aritie s and Diffe renc es b etwe en
ANO VA , R eg re ss ion , and D isc rim ina nt
An aly sis
Table 18.1
Differ en ces
Nature of the
de pen den t Me tr ic Me tr ic Categor ic al
varia bl es
Nature of the
indep en den t Ca te goric al Me tr ic Metr ic
varia bl es
18-5
© 2007 Prentice Hall
Discr imina nt Ana lysis
Discr imi nan t an al ys is is a technique for analyzing data
when the criterion or dependent variable is categorical and
the predictor or independent variables are interval in nature.
The objectives of discriminant analysis are as follows:
Development of discri min ant fun ct io ns, or linear
X2 G1
1 1 2 2
G2
1 1 11 2
1 1 1 1 2
1
2 2
2 22
1 2
21
1 22
22
G1
G2 X1
18-8
© 2007 Prentice Hall
Discr iminant An alysis Mo del
The dis cr imi nant an alysis model involves linear combinations of
the following form:
D = b0 + b1X1 + b2X2 + b3X3 + . . . + bkXk
Where:
D = discriminant score
b 's = discriminant coefficient or weight
X 's = predictor or independent variable
The coefficients, or weights (b), are estimated so that the groups differ
as much as possible on the values of the discriminant function.
This occurs when the ratio of between-group sum of squares to within-
group sum of squares for the discriminant scores is at a maximum.
18-9
© 2007 Prentice Hall
St ati stics A sso ciated w ith
Discr iminant An alysis
Ca nonical co rre lat io n. Canonical correlation
measures the extent of association between the
discriminant scores and the groups. It is a measure of
association between the single discriminant function and
the set of dummy variables that define the group
membership.
Ce ntr oid. The centroid is the mean values for the
discriminant scores for a particular group. There are as
many centroids as there are groups, as there is one for
each group. The means for a group on all the functions
are the group centroids.
Cl as sif icat ion mat rix . Sometimes also called
confusion or prediction matrix, the classification matrix
contains
© 2007 Prentice Hall the number of correctly classified and 18-10
Stati sti cs Asso ciated w ith
Discr imina nt Ana lysis
Discr imi nan t function co ef fici en ts. The discriminant
function coefficients (unstandardized) are the multipliers
of variables, when the variables are in the original units
of measurement.
Discr imi nan t sco res. The unstandardized coefficients
are multiplied by the values of the variables. These
products are summed and added to the constant term to
obtain the discriminant scores.
Eig en val ue . For each discriminant function, the
Eigenvalue is the ratio of between-group to within-group
sums of squares. Large Eigenvalues imply superior
functions.
18-11
© 2007 Prentice Hall
St ati stics A sso ciated w ith
Discr iminant An alysis
F va lu es an d the ir si gni fican ce . These are calculated
from a one-way ANOVA, with the grouping variable
serving as the categorical independent variable. Each
predictor, in turn, serves as the metric dependent variable
in the ANOVA.
Group mean s and gro up st an dar d de vi at ions.
These are computed for each predictor for each group.
Po ole d wi thin -gro up co rrel at io n ma tr ix. The pooled
within-group correlation matrix is computed by averaging
the separate covariance matrices for all the groups.
18-12
© 2007 Prentice Hall
Stati st ics A sso ciated with
Di sc ri mi nant Ana ly si s
St anda rdiz ed d is crimin an t fu nct ion coefficient s. The
standardized discriminant function coefficients are the discriminant
function coefficients and are used as the multipliers when the
variables have been standardized to a mean of 0 and a variance of 1.
St ruc tu re corr elat ions. Also referred to as discriminant loadings,
the structure correlations represent the simple correlations between
the predictors and the discriminant function.
Tot al c orre la tion ma trix . If the cases are treated as if they were
from a single sample and the correlations computed, a total
correlation matrix is obtained.
Wilks 'λ . Sometimes also called the U statistic, Wilks'λ for each
predictor is the ratio of the within-group sum of squares to the total
sum of squares. Its value varies between 0 and 1. Large values
of λ (near 1) indicate that group means do not seem to be different.
Small values ofλ (near 0) indicate that the group means seem to be
different.
18-13
© 2007 Prentice Hall
Co nduc ti ng Di scrim inant
An al ysi s
Fig. 18.2
1 1 50.2 5 8 3 43 M (2)
2 1 70.3 6 7 4 61 H (3)
3 1 62.9 7 5 6 52 H (3)
4 1 48.5 7 5 5 36 L (1)
5 1 52.7 6 6 4 55 H (3)
6 1 75.0 8 7 5 68 H (3)
7 1 46.2 5 3 3 62 M (2)
8 1 57.0 2 4 6 51 M (2)
9 1 64.1 7 5 4 57 H (3)
10 1 68.1 7 6 5 45 H (3)
11 1 73.4 6 7 5 44 H (3)
12 1 71.9 5 8 4 64 H (3)
13 1 56.2 1 8 6 54 M (2)
18-16
© 200714Prentice Hall1 49.3 4 2 3 56 H (3)
Info rm at ion on Resort Vis it s: Ana lys is
Sam pl e
Table 18.2, cont.Annual Attitud e Importance Ho use ho ld Ag e of
Amou nt
Res ort Famil y Tow ar d A ttached Si ze Hea d o f
Sp ent on N o. Visi t
Income Travel to Fam ily Ho use hol d Famil y
($ 000) V ac ation
Vacati on
16 2 32.1 5 4 3 58 L (1)
17 2 36.2 4 3 2 55 L (1)
18 2 43.2 2 5 2 57 M (2)
19 2 50.4 5 2 4 37 M (2)
20 2 44.1 6 6 3 42 M (2)
21 2 38.3 6 6 2 45 L (1)
22 2 55.0 1 2 2 57 M (2)
23 2 46.1 3 5 3 51 L (1)
24 2 35.0 6 4 5 64 L (1)
25 2 37.3 2 7 4 54 L (1)
26 2 41.8 5 1 3 56 M (2)
27 2 57.0 8 3 2 36 M (2)
28 2 33.4 6 8 2 50 L (1)
29
© 2007 Prentice Hall
2 37.5 3 2 3 48 L (1) 18-17
Info rm ati on on Re sort Vis it s:
1 1 50.8 4 7 3 45 M(2)
2 1 63.6 7 4 7 55 H (3)
3 1 54.0 6 7 4 58 M(2)
4 1 45.0 5 4 3 60 M(2)
5 1 68.0 6 6 6 46 H (3)
6 1 62.1 5 6 3 56 H (3)
7 2 35.0 4 3 4 54 L (1)
8 2 49.6 5 3 5 39 L (1)
9 2 39.4 6 5 3 44 H (3)
18-18
© 2007 10
Prentice Hall 2 37.0 2 6 5 51 L (1)
Con duc ting D isc rimi na nt A na lysi s
Esti ma te t he Di sc ri mi nant
Func tion C oef fic ient s
18-19
© 2007 Prentice Hall
Ana lys is
Table
18.4
GROUP MEANS
VISIT INCOME TRAVEL VACATION HSIZE AGE
INCOME 1.00000
TRAVEL 0.19745 1.00000
VACATION 0.09148 0.08434 1.00000
HSIZE 0.08887 -0.01681 0.07046 1.00000
AGE - 0.01431 -0.19709 0.01742 -0.04301 1.00000
* marks t he 1 canon ical d isc riminan t fun ctio ns remaining in the analy sis .
Stand ard Cano nical Dis criminan t Funct ion Co eff icients
FUN C 1
INCO ME 0.74 30 1
TR AV EL 0.09 61 1
VACA TI ON 0.23 32 9
HSI ZE 0.46 91 1
AG E 0.20 92 2
Structure Matri x:
Po oled within -gr oup s c or relations be twee n dis cr iminating vari ables & canon ical d isc riminan t fu nctio ns
(variables or dered by siz e o f co rr elation within fu nctio n)
FUN C 1
INCO ME 0.82 20 2
HSI ZE 0.54 09 6
VACA TI ON 0.34 60 7
TR AV EL 0.21 33 7
AG E 0.16 35 4 Co nt.
18-21
© 2007 Prentice Hall
Res ults of Tw o-G ro up Di sc rim ina nt
Analys is
Table 18.4, cont.
Gro up FUNC 1
1 1.2 911 8
2 -1.29 11 8
Cla ssifica tion result s fo r cases sel ect ed for use in ana ly sis
Pr edict ed Gro up Memb ersh ip
Ac tua l Grou p No . of Ca ses 1 2
Gro up 1 15 12 3
80 .0 % 20 .0 %
Gro up 2 15 0 15
0. 0% 10 0. 0%
Per cent of gro uped ca ses co rrect ly cla ssifi ed : 90. 00 %
Co nt.
18-22
© 2007 Prentice Hall
Res ults of Tw o-G ro up Di sc rim ina nt
Analys is
Table 18.4, cont.
18-23
© 2007 Prentice Hall
Con duc ting Di scr imi na nt A na lysi s
De term ine t he S ign if icanc e o f
Di sc ri mi nant Func tio n
The null hypothesis that, in the population, the means of all
discriminant functions in all groups are equal can be
statistically tested.
In SPSS this test is based on Wilks'λ . If several functions
are tested simultaneously (as in the case of multiple
discriminant analysis), the Wilks' λ statistic is the product of
the univariate for each function. The significance level is
estimated based on a chi-square transformation of the
statistic.
If the null hypothesis is rejected, indicating significant
discrimination, one can proceed to interpret the results.
18-24
© 2007 Prentice Hall
Int erpre t t he R esul ts
The interpretation of the discriminant weights, or coefficients, is similar to
that in multiple regression analysis.
Given the multicollinearity in the predictor variables, there is no
unambiguous measure of the relative importance of the predictors in
discriminating between the groups.
With this caveat in mind, we can obtain some idea of the relative
importance of the variables by examining the absolute magnitude of the
standardized discriminant function coefficients.
Some idea of the relative importance of the predictors can also be
obtained by examining the structure correlations, also called canonical
loadings or discriminant loadings. These simple correlations between each
predictor and the discriminant function represent the variance that the
predictor shares with the function.
Another aid to interpreting discriminant analysis results is to develop a
Ch arac teri sti c prof il e for each group by describing each group in terms
of the group means for the predictor variables.
18-25
© 2007 Prentice Hall
Conduc ting Di sc rim inant Ana lys is
Asses s Vali dit y of Di scr im ina nt A nal ys is
Many computer programs, such as SPSS, offer a leave-one-out
cross-validation option.
The discriminant weights, estimated by using the analysis
sample, are multiplied by the values of the predictor variables in
the holdout sample to generate discriminant scores for the
cases in the holdout sample. The cases are then assigned to
groups based on their discriminant scores and an appropriate
decision rule. The hit rat io, or the percentage of cases
correctly classified, can then be determined by summing the
diagonal elements and dividing by the total number of cases.
It is helpful to compare the percentage of cases correctly
classified by discriminant analysis to the percentage that would
be obtained by chance. Classification accuracy achieved by
discriminant analysis should be at least 25% greater than that
obtained by chance.
18-26
© 2007 Prentice Hall
Re sul ts o f Three -Gro up D isc rimi na nt
Analy sis
Table 18.5
Grou p Me an s
AM OUN T INCO ME TR AV EL VACA TI ON HSI ZE AG E
1 38.57 000 4. 500 00 4.70 000 3 .100 00 50 .3000 0
2 50.11 000 4. 000 00 4.20 000 3.40 00 0 49 .5000 0
3 64.97 000 6. 100 00 5.90 000 4 .200 00 56 .0000 0
Total 51.21 667 4. 866 67 4.93 333 3 .566 67 51 .9333 3
* marks t he two canon ical disc rimin ant fu nctio ns remainin g in the analy si s.
St and ard ize d Ca no ni cal Disc rimi na nt Funct ion Coeff ici ent s
FUN C 1 FUN C 2
INCO ME 1.04 740 -0 .420 76
TRA VEL 0.33 991 0.76 851
VACAT ION -0 .141 98 0.53 354
HSI ZE -0 .163 17 0.12 932 Co nt.
AG E 0.49 474 0.52 447
18-28
© 2007 Prentice Hall
Resul ts of T hree -Group Discr imi na nt
Ana lysi s
Table 18.5, cont.
Structure Matr ix:
Pooled wi th in- gr oup s co rr el atio ns betwe en d iscr iminating variab les and canon ical
dis crimin ant functio ns (v aria ble s o rde re d b y si ze o f correl atio n w ithin f unctio n)
FUN C 1 FUN C 2
INCO ME 0.85 55 6* -0 .278 33
HSI ZE 0.19 31 9* 0.07 74 9
VACATI ON 0.21 93 5 0.58 82 9*
TR AV EL 0.14 89 9 0.45 36 2*
AG E 0.16 57 6 0.34 07 9*
Un stand ard ize d cano nical d is cri mina nt functio n co eff icients
FUN C 1 FUN C 2
INCO ME 0.15 42 658 -0 .619 714 8E- 01
TR AV EL 0.18 67 977 0.42 23 430
VACATI ON -0 .695 226 4E- 01 0.26 12 652
HSI ZE -0 .126 533 4 0.10 02 796
AG E 0 .592 805 5E- 01 0.62 84 206 E-0 1
(co nstant) -1 1.09 442 -3 .791 600
Cano nica l d iscri minant f unctio ns eval uated a t gro up mea ns (g ro up centro ids)
Gro up FUN C 1 FUN C 2
1 -2 .041 00 0.41 84 7
2 -0 .404 79 -0 .658 67
3 2 .445 78 0.24 02 0 Co nt.
18-29
© 2007 Prentice Hall
Re sul ts o f Three -Gro up D isc rimi na nt
Analy sis
Table 18.5, cont.
Classi fic at ion Re sul ts:
Pr edi cted Gro up M embers hip
Actual Grou p No. o f Cas es 1 2 3
Gr oup 1 10 9 1 0
90 .0% 10 .0% 0.0%
Gr oup 2 10 1 9 0
10 .0% 90 .0% 0.0%
Gr oup 3 10 0 2 8
0.0% 20 .0% 80 .0%
Perc ent of gr ouped cases c orr ectly clas sif ied: 8 6.67 %
Classi fic at ion resul ts for cases n ot sel ect ed for use in the
ana lysi s
Pr edi cted Gro up M embers hip
Actual G ro up No. o f C ases 1 2 3
Gro up 1 4 3 1 0
75 .0% 25 .0% 0.0%
Gro up 2 4 0 3 1
0.0% 75 .0% 25 .0%
Gro up 3 4 1 0 3
25 .0% 0.0% 75 .0%
Perc ent of gr ouped cases c orr ectly clas sif ied: 7 5.00 % 18-30
© 2007 Prentice Hall
Al l-Groups S ca tter gram
Fig. 18.3
Acros s: Functi on
1
Down : Fu nct ion 2
4.
0 1 1
1 *1 3
23 3 *3 3
1 1 12 * 3 3
0. 1 1 2 2
3
0 1 2 2
2
-4.0
13
13
13 Acr oss: Function 1
8.0 Down : Fu ncti on 2
13
13 * Ind icates a
13
group centro id
13
4.0 113
1 1 2
31 1 2 2 3 3
*1 1 1 2 2 2 2 3 3 *
1 1 2 * 2 2 3
0.0 12 1 2 2 3 3
1 1 22 2 2 3 3
1 121 2 2 2 3
2 3 3
-4.0 1 21 2 2
21 1 2 2 2 3
1 1 21 2 2 3 3
1 12 1 2 2 2 2 3 3
1 1 2 2 2 3
-8.0 1 1 1 22 2 2 3 3
-8.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 18-32
© 2007 Prentice Hall
St epw ise Discr imi na nt Ana lys is
Stepwise discriminant analysis is analogous to stepwise
multiple regression (see Chapter 17) in that the predictors
are entered sequentially based on their ability to
discriminate between the groups.
An F ratio is calculated for each predictor by conducting a
univariate analysis of variance in which the groups are
treated as the categorical variable and the predictor as the
criterion variable.
The predictor with the highest F ratio is the first to be
selected for inclusion in the discriminant function, if it
meets certain significance and tolerance criteria.
A second predictor is added based on the highest adjusted
or partial F ratio, taking into account the predictor already
selected.
18-33
© 2007 Prentice Hall
Stepw ise D iscr imi na nt Ana lys is
Each predictor selected is tested for retention based on its
association with other predictors selected.
The process of selection and retention is continued until all
predictors meeting the significance criteria for inclusion and
retention have been entered in the discriminant function.
The selection of the stepwise procedure is based on the
optimizing criterion adopted. The Mah ala no bi s
pr oce du re is based on maximizing a generalized measure
of the distance between the two closest groups.
The order in which the variables were selected also
indicates their importance in discriminating between the
groups.
18-34
© 2007 Prentice Hall
Th e Logi t Mo del
18-35
© 2007 Prentice Hall
Bi na ry Log it Mo del Formul ati on
The probability of success may be modeled using the logit model as:
P
Or log = a +a X +a X +... +a X
1 − P
e 0 1 1 2 2 k k
P = n
Or log ∑a X
1 − P
e i i
i= 0
18-36
© 2007 Prentice Hall
Mo del F ormu latio n
exp( ∑ a X )
P =
i i
i =0
1 + exp(
k
∑ a X
i =0
i i )
Where:
P = Probability of success
Xi = Independent variable i
ai = parameter to be estimated.
18-37
© 2007 Prentice Hall
Prop erti es of the Lo git Model
18-38
© 2007 Prentice Hall
Es ti mati on a nd Mod el F it
The estimation procedure is called the maximum
likelihood method.
Fit: Cox & Snell R Square and Nagelkerke R Square.
Both these measures are similar to R2 in multiple
regression.
The Cox & Snell R Square can not equal 1.0, even if the
fit is perfect
This limitation is overcome by the Nagelkerke R Square.
Compare predicted and actual values of Y to determine
the percentage of correct predictions.
18-39
© 2007 Prentice Hall
Signi fi can ce Tes ting
18-40
© 2007 Prentice Hall
Inte rpre tati on o f Co eff ici ents
18-41
© 2007 Prentice Hall
Ex plai ning Bran d Lo yal ty
Table 18.6
No. Loyalty Brand Product Shopping
1 1 4 3 5
2 1 6 4 4
3 1 5 2 4
4 1 7 5 5
5 1 6 3 4
6 1 3 4 5
7 1 5 5 5
8 1 5 4 2
9 1 7 5 4
10 1 7 6 4
11 1 6 7 2
12 1 5 6 4
13 1 7 3 3
14 1 5 1 4
15 1 7 5 5
16 0 3 1 3
17 0 4 6 2
18 0 2 5 2
19 0 5 2 4
20 0 4 1 3
21 0 3 3 4
22 0 3 4 5
23 0 3 6 3
24 0 4 4 2
25 0 6 3 6
26 0 3 6 3
27 0 4 3 2
28 0 3 5 2
29 0 5 5 3
30 0 1 3 2
18-42
© 2007 Prentice Hall
Resu lts of Log isti c Re gres si on
Table 18.7
M odel Su m m ary
18-43
© 2007 Prentice Hall
Resu lts of Log isti c Regres si on
Table 18.7, cont.
Cl ass if icati on Ta ble a
Predicted
18-45
© 2007 Prentice Hall
SPSS Wi ndo ws: Two- gro up
Dis cri mi nant
1. Select ANALYZE from the SPSS menu bar.
2. Click CLASSIFY and then DISCRIMINANT.
3. Move “visit” in to the GROUPING VARIABLE box.
4. Click DEFINE RANGE. Enter 1 for MINIMUM and 2 for MAXIMUM. Click
CONTINUE.
5. Move “income,” “travel,” “vacation,” “hsize,” and “age” in to the
INDEPENDENTS box.
6. Select ENTER INDEPENDENTS TOGETHER (default option)
7. Click on STATISTICS. In the pop-up window, in the DESCRIPTIVES
box check MEANS and UNIVARIATE ANOVAS. In the MATRICES box
check WITHIN-GROUP CORRELATIONS. Click CONTINUE.
8. Click CLASSIFY.... In the pop-up window in the PRIOR PROBABILITIES
box check ALL GROUPS EQUAL (default). In the DISPLAY box check
SUMMARY TABLE and LEAVE-ONE-OUT CLASSIFICATION. In the USE
COVARIANCE MATRIX box check WITHIN-GROUPS. Click CONTINUE.
9. Click OK.
18-46
© 2007 Prentice Hall
SP SS Wi ndow s: Logi t An alysi s
1. Select ANALYZE from the SPSS menu bar.
6. Click OK.
18-47
© 2007 Prentice Hall