You are on page 1of 39

SPSS - ANOVA

Miquel A. Belmonte
Hospital General de Castelln
Espaa, 2000

Eleccin de tests estadsticos


Varbl. dependientes
>1

1
Varbl.
Indep.
1
>1

CATEG

CUANT

CATEG

X2

Oneway

CUANT

Student T

Correlac

CATEG

Loglinear

ANOVA

CUANT

Reg. Logist Reg.Mult

CATEG CUANT
Manova

Manova

ANOVA - conceptos bsicos


Influencia de una o varias variables categricas
(factores) sobre una variable dependiente cuantitativa
Valora efectos principales de factores e interacciones
de stos entre s
Admite una o ms covariables de control, de tipo
cuantitativo
Estudia reduccin de variabilidad (suma de cuadrados)
Test paramtrico: Compara las medias de los subgrupos
formados para cada factor

Condiciones
Igualdad de varianzas
Distribucin paramtrica
a
Levene's Test of Equality of Error Variances

Dependent Variable: Current salary


F
10,374

df1
19

df2
454

Sig.
,000

Tests the null hypothesis that the error variance of the


dependent variable is equal across groups.
a. Design: Intercept+AGE+SEXRACE+JOBCAT+SEXRACE
* JOBCAT

ANOVA - estadsticos
SS total = SS efectos principales +
SS interacciones orden 2 y sucesivas +
SS residuos
SS (factor o interaccin)
MS =
DF (grados de libertad)
MS (factor o interaccin)
F=
MS de los residuos

ANOVA - Output SPSS


Source of Variation
Covariates
AGE
Main Effects
JOBCAT
SEXRACE
2-Way Interactions
JOBCAT
SEXRACE

Sum of
Squares

Mean
Square

DF

601239435
601239435

1 601239435,132
1 601239435,132

54,882
54,882

,000
,000

11677985838
7428829687
709773003

7 1668283691,12
4 1857207421,83
3 236591000,879

152,284
169,529
21,596

,000
,000
,000

39769437,169
39769437,169

3,630
3,630

,000
,000

16 787336298,143

71,869

,000

318155497
318155497

8
8

Explained

12597380770

Residual

4885977728

446

10955107,014

17483358499

462

37842767,313

Total

Sum of Squares
Mean Square = _______________
DF

Sig
of F

F = _____________________
MS Factor explained
MS Residual

Modelos de ANOVA
ANOVA factorial general
ANOVA multivariado: MANOVA
ANOVA de medidas repetidas

ANOVA
Modelo Lineal General Factorial

Definicin del modelo

Opciones

Contrastes
Permiten comparar
niveles o categoras
entre s, dentro de cada
factor considerado.

Test Results
Dependent Variable: Current salary
Source
Contrast
Error

Sum of
Squares
1,09E+09
5,47E+09

df
6
453

Mean
Square
1,82E+08
12068299

F
15,050

Sig.
,000

Eta
Squared
,166

Profile Plots
Muestran estimaciones de
Media Marginal
para cada subgrupo formado.

40000

Estimated Marginal Means of Current salary


24000
22000

30000

18000

Mean of Current salary

Estimated Marginal Means

20000

16000
14000
12000
10000

20000

10000

0
Clerical

8000
White males

Minority males

Sex & race classification

White females

Minority females

Security officer
Office trainee

Employment category

Exempt employee

College trainee

MBA trainee

Technical

Grfico de distribucin
Spread vs. Level Plot of Current salary
12000

Spread (Standard Deviation)

10000

8000

6000

4000

2000
0
0

10000

20000

30000

Level (Mean)
Groups: Sex & race classification * Employment category

40000

ANOVA
Modelo Lineal General Factorial
SYNTAX
UNIANOVA
salnow BY sexrace jobcat WITH age
/METHOD = SSTYPE(3)
/INTERCEPT = INCLUDE
/PRINT = DESCRIPTIVE ETASQ HOMOGENEITY
/CRITERIA = ALPHA(.05)
/DESIGN = age sexrace jobcat sexrace*jobcat .

Descriptive Statistics
Dependent Variable: Current salary
Sex & race classification
White males

Estadsticas
Descriptivas

Minority males

White females

Minority females

Total

Employment category
Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Technical
Total
Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Total
Clerical
Office trainee
College trainee
Exempt employee
MBA trainee
Total
Clerical
Office trainee
Total
Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Technical
Total

Mean
13057,65
13092,23
12471,43
24916,97
25570,71
26916,67
36691,67
17790,16
11752,57
11080,00
12272,31
31400,00
31880,00
26500,00
12898,44
9890,12
10501,78
18040,57
19660,00
23250,00
10682,72
9258,75
9090,00
9225,00
11134,82
11136,41
12375,56
23901,07
25595,62
26100,00
36691,67
13767,83

Std.
Deviation
3251,46
3839,48
663,50
5321,94
7152,95
3003,47
10543,45
8132,26
2561,80
1086,75
1025,17
,
11483,41
,
5223,95
2839,32
1894,68
3171,27
4299,21
,
3204,76
1719,52
972,77
1588,95
3196,57
2732,60
845,85
5695,15
7364,40
2661,06
10543,45
6830,26

N
75
35
14
33
28
3
6
194
35
12
13
1
2
1
64
85
81
7
2
1
176
32
8
40
227
136
27
41
32
5
6
474

ANOVA Output
Tests of Between-Subjects Effects
Dependent Variable: Current salary

Source
Corrected Model
Intercept
AGE
SEXRACE
JOBCAT
SEXRACE * JOBCAT
Error
Total
Corrected Total

Type III
Sum of
Squares
1,66E+10a
9,28E+09
2,06E+08
4,06E+08
5,50E+09
3,20E+08
5,47E+09
1,12E+11
2,21E+10

df
20
1
1
3
6
10
453
474
473

a. R Squared = ,752 (Adjusted R Squared = ,741)

Mean
Square
8,30E+08
9,28E+09
2,06E+08
1,35E+08
9,17E+08
31952108
12068299

F
68,774
768,594
17,075
11,205
75,953
2,648

Sig.
,000
,000
,000
,000
,000
,004

Eta
Squared
,752
,629
,036
,069
,501
,055

ANOVA factorial
Simple

Diseos factoriales de modelos saturados


Mtodos:
nico: todos los elementos concurrentemente
Jerrquico: covariables - factores- interaccin
Experimental: factores- interaccin

General

Diseos factoriales de modelos no saturados


Permite especificar con ms flexibilidad el modelo a
utilizar y variedad de estadsticos

ONEWAY
Caso particular de ANOVA de un factor
Una sola variable dependiente cuantitativa
Un factor con varias categoras
ONEWAY, pero no ANOVA, produce:

Contrastes
Comparaciones mltiples
Pruebas de tendencia
Test homogeneidad de varianza

ONEWAY

Resultados Oneway
Listado descriptivo
Descriptives
Current salary

N
Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Technical
Total

227
136
27
41
32
5
6
474

Mean
11134,82
11136,41
12375,56
23901,07
25595,63
26100,00
36691,67
13767,83

Std.
Deviation
3196,57
2732,60
845,85
5695,15
7364,40
2661,06
10543,45
6830,26

Std. Error
212,16
234,32
162,78
889,43
1301,86
1190,06
4304,35
313,72

95% Confidence Interval


for Mean
Lower
Upper
Bound
Bound
10716,75
11552,89
10673,00
11599,82
12040,95
12710,16
22103,46
25698,68
22940,47
28250,78
22795,86
29404,14
25626,99
47756,34
13151,36
14384,29

Minimum
6300
7260
9720
13764
15480
23250
26700
6300

Maximum
26750
32000
14100
36500
41500
30000
54000
54000

ONEWAY - Output SPSS


Analysis of Variance

Source
Between Groups

D.F.

Sum of
Squares

Mean
Squares

F
Ratio

F
Prob.

11168758807

2792189702

202,5184

,0000

Unweighted Linear Term


Weighted Linear Term
Deviation from Linear
Within Groups

1
1
3
458

9966174401
8987023511
2181735296
6314599691

9966174401
8987023511
727245098,6
13787335,57

722,8499
651,8318
52,7473

,0000
,0000
,0000

Total

462

17483358499
ANOVA

Current salary

Between Groups
Within Groups
Total

Sum of
Squares
1,52E+10
6,90E+09
2,21E+10

df
6
467
473

Mean
Square
2,53E+09
14772477

F
171,128

Sig.
,000

ONEWAY - Comparaciones Mltiples


Multiple Range Tests:

Scheffe test with significance level ,05

The difference between two means is significant if


MEAN(J)-MEAN(I) >= 2625,5795 * RANGE * SQRT(1/N(I) + 1/N(J))
with the following value(s) for RANGE: 4,37

(*) Indicates significant differences which are shown in the lower triangle
C O S
E
l f e C x
e f c o e
r i u l m
Mean
11134,8194
11136,4118
12375,5556

JOBCAT
Clerical
Office t
Security

23901,0732

College

25595,6250

Exempt e

* * *
* * *

Oneway - Contrastes
Current salary

Employment category
Tukey HSDa,b Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Technical
Sig.
Scheffea,b
Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Technical
Sig.

N
227
136
27
41
32
5
6
227
136
27
41
32
5
6

Subset for alpha = .05


1
2
3
11134,82
11136,41
12375,56
23901,07
25595,63
26100,00
36691,67
,976
,708
1,000
11134,82
11136,41
12375,56
23901,07
25595,63
26100,00
36691,67
,993
,876
1,000

Means for groups in homogeneous subsets are displayed.


a. Uses Harmonic Mean Sample Size = 14,859.
b. The group sizes are unequal. The harmonic mean of the group sizes is used.
Type I error levels are not guaranteed.

Oneway - Means Plot


40000

Mean of Current salary

30000

20000

10000

0
Clerical

Security officer
Office trainee

Employment category

Exempt employee

College trainee

MBA trainee

Technical

Tipos de Anlisis de Varianza


Nmero de Varbl. independientes
1

Condiciones
previas

Paramtrico

Oneway

No
Paramtrico

KruskalWallis

>1
independ
ANOVA

>1
relacionada

Friedman

Kruskal-Wallis
Oneway no paramtrico
Tests no paramtricos - K muestras independientes

Kruskal-Wallis
Estadsticos descriptivos

Ranks

Current salary

Employment category
Clerical
Office trainee
Security officer
College trainee
Exempt employee
MBA trainee
Technical
Total

N
227
136
27
41
32
5
6
474

Mean Rank
187,70
194,82
278,98
422,15
427,33
438,30
460,83

Test Statisticsb

N
Median
Chi-Square
df
Asymp. Sig.

Kruskal-Wallis

Current
salary
474
11550,00
135,133a
6
,000

Prueba de la Mediana

a. 4 cells (,0%) have expected frequencies less than 5.


The minimum expected cell frequency is 2,5.
b. Grouping Variable: Employment category

Frequencies

Current salary

> Median
<= Median

Clerical
80
147

Office
trainee
48
88

Employment category
Security
College
Exempt
officer
trainee
employee
25
41
32
2
0
0

MBA
trainee
5
0

Technical
6
0

Kruskal-Wallis
Contrastes

Estadsticos de contrasteb,c

Chi-cuadrado
gl
Sig. asintt.
Sig. Monte
Carlo

Current
salary
208,357
6
,000
,000a
,000
,000

Sig.
Intervalo de
Lmite inferior
confianza al
Lmite superior
99%
a. Basado en 10000 tablas muestrales con semilla de inicio
2000000.
b. Prueba de Kruskal-Wallis
c. Variable de agrupacin: Employment category

Friedman
Anova de dos vas para muestras apareadas
Friedman es el test
bsico, que compara
los rangos entre K
variables relacionadas
entre s. Se basa en X2

Kendall W es un test de concordancia donde cada variable


es un sujeto de estudio y cada caso es un juez.
Cochran Q se usa cuando todas las variables independientes
son dicotmicas

Friedman
Resultados
Descriptive Statistics
N
HBA1_BASAL
HBA1_SIST1
HBA1_SIST2
HBA1_SIST3

30
30
30
30

Mean
7,657
7,660
7,250
7,117

Ranks
Mean Rank
HBA1_BASAL
2,80
HBA1_SIST1
3,15
HBA1_SIST2
2,28
HBA1_SIST3
1,77

Std.
Deviation
1,748
1,123
1,279
1,372

Minimum
4,8
6,2
5,8
5,2

Maximum
11,8
10,3
11,6
10,5

Test Statistics
N
Kendall's Wa
Chi-Square
df
Asymp. Sig.

30
,229
20,573
3
,000

a. Kendall's Coefficient of Concordance

ANOVA - otros modelos


ANOVA multivariado: MANOVA
Ms de una variable dependiente
Gran complejidad de modelos

ANOVA medidas repetidas


La variable dependiente se mide en ms de una
ocasin para cada sujeto
Multivariante complejo

MANOVA

Multivariate Testsc
Effect
Intercept

Manova Resultados

EDAD

T_EVOLUC

DIETA_KC

DOSIS_T1

Between-Subjects Factors
Value
Label
Num.inyecc
diarias
Nivel
educacion
diabetologica

Sexo

3
4
3
5
6
7
8
9
1
2

Hombre
Mujer

INYECC_D

N
9
21
2
3
5
5
7
8
12
18

CALIF_ED

SEXO

INYECC_D * CALIF_ED

INYECC_D * SEXO

CALIF_ED * SEXO

INYECC_D * CALIF_ED *
SEXO

Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root

Value
,758
,242
3,138
3,138
,370
,630
,586
,586
,118
,882
,134
,134
,240
,760
,316
,316
,127
,873
,146
,146
,212
,788
,268
,268
,703
,404
1,211
,924
,210
,790
,266
,266
,313
,687
,456
,456
,155
,845
,184
,184
,367
,654
,495
,416
,000
1,000
,000
,000

F
15,691a
15,691a
15,691a
15,691a
2,932a
2,932a
2,932a
2,932a
,669a
,669a
,669a
,669a
1,578a
1,578a
1,578a
1,578a
,728a
,728a
,728a
,728a
1,341a
1,341a
1,341a
1,341a
1,192
1,147a
1,090
2,033b
1,332a
1,332a
1,332a
1,332a
2,279a
2,279a
2,279a
2,279a
,918a
,918a
,918a
,918a
,825
,787a
,743
1,525b
,a
,a
,a
,000a

Hypothesis
df
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
10,000
10,000
10,000
5,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
2,000
6,000
6,000
6,000
3,000
,000
,000
,000
2,000

Error df
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
22,000
20,000
18,000
11,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
10,000
22,000
20,000
18,000
11,000
,000
10,500
2,000
9,000

Sig.
,001
,001
,001
,001
,100
,100
,100
,100
,534
,534
,534
,534
,254
,254
,254
,254
,507
,507
,507
,507
,305
,305
,305
,305
,348
,379
,419
,152
,307
,307
,307
,307
,153
,153
,153
,153
,430
,430
,430
,430
,563
,590
,622
,263
,
,
,
1,000

a. Exact statistic
b. The statistic is an upper bound on F that yields a lower bound on the significance level.
c. Design: Intercept+EDAD+T_EVOLUC+DIETA_KC+DOSIS_T1+INYECC_D+CALIF_ED+SEXO+INYECC_D *
CALIF_ED+INYECC_D * SEXO+CALIF_ED * SEXO+INYECC_D * CALIF_ED * SEXO

Eta
Squared
,758
,758
,758
,758
,370
,370
,370
,370
,118
,118
,118
,118
,240
,240
,240
,240
,127
,127
,127
,127
,212
,212
,212
,212
,351
,364
,377
,480
,210
,210
,210
,210
,313
,313
,313
,313
,155
,155
,155
,155
,184
,191
,198
,294
,
,
,
,000

ANOVA medidas repetidas

Se define un factor intra-sujetos con


el nmero de categoras o niveles adecuado
al nmero de mediciones realizadas

Se especifican las variables de las mediciones


realizadas como grupos del factor intra-sujetos creado.
Pueden definirse tambin covariables de correccin y
factores de agrupamiento entre-sujetos.

Within-Subjects Factors
Measure: MEASURE_1
FACTOR1
1
2
3
4

Dependent
Variable
HBA1_BAS
HBA1_S1
HBA1_S2
HBA1_S3

Resultados - Efectos
intrasujetos
Tests of Within-Subjects Effects
Measure: MEASURE_1

Source
FACTOR1

Las pruebas de efectos


entre sujetos icnluyen
diversos tests.
Se determinan tambin
estadsticos para las
interacciones con el factor
estudiado.

Sphericity Assumed
Greenhouse-Geisser
Huynh-Feldt
Lower-bound
FACTOR1 * EDAD
Sphericity Assumed
Greenhouse-Geisser
Huynh-Feldt
Lower-bound
FACTOR1 * T_EVOLUC Sphericity Assumed
Greenhouse-Geisser
Huynh-Feldt
Lower-bound
FACTOR1 * SEXO
Sphericity Assumed
Greenhouse-Geisser
Huynh-Feldt
Lower-bound
Error(FACTOR1)
Sphericity Assumed
Greenhouse-Geisser
Huynh-Feldt
Lower-bound

Type III
Sum of
Squares
3,343
3,343
3,343
3,343
1,962
1,962
1,962
1,962
1,594
1,594
1,594
1,594
2,583
2,583
2,583
2,583
33,039
33,039
33,039
33,039

df
3
1,851
2,216
1,000
3
1,851
2,216
1,000
3
1,851
2,216
1,000
3
1,851
2,216
1,000
78
48,115
57,618
26,000

Mean
Square
1,114
1,807
1,509
3,343
,654
1,060
,885
1,962
,531
,862
,719
1,594
,861
1,396
1,165
2,583
,424
,687
,573
1,271

F
2,631
2,631
2,631
2,631
1,544
1,544
1,544
1,544
1,255
1,255
1,255
1,255
2,033
2,033
2,033
2,033

Sig.
,056
,086
,075
,117
,210
,225
,221
,225
,296
,292
,295
,273
,116
,145
,136
,166

ANOVA medidas repetidas

Pruebas multivariadas
Multivariate Testsb
Effect
FACTOR1

FACTOR1 * EDAD

FACTOR1 * T_EVOLUC

FACTOR1 * SEXO

Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root

a. Exact statistic
b.
Design: Intercept+EDAD+T_EVOLUC+SEXO
Within Subjects Design: FACTOR1

Value
,234
,766
,306
,306
,119
,881
,136
,136
,240
,760
,316
,316
,107
,893
,120
,120

F
2,446a
2,446a
2,446a
2,446a
1,086a
1,086a
1,086a
1,086a
2,526a
2,526a
2,526a
2,526a
,960a
,960a
,960a
,960a

Hypothesis
df
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000
3,000

Error df
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000
24,000

Sig.
,088
,088
,088
,088
,374
,374
,374
,374
,081
,081
,081
,081
,428
,428
,428
,428

ANOVA medidas repetidas

Contrastes intra-sujetos
Tests of Within-Subjects Contrasts
Measure: MEASURE_1

Source
FACTOR1

FACTOR1 * EDAD

FACTOR1 * T_EVOLUC

FACTOR1 * SEXO

Error(FACTOR1)

FACTOR1
Linear
Quadratic
Cubic
Linear
Quadratic
Cubic
Linear
Quadratic
Cubic
Linear
Quadratic
Cubic
Linear
Quadratic
Cubic

Type III
Sum of
Squares
2,609
,532
,202
,504
1,437
2,133E-02
1,680E-02
1,232
,346
,683
1,782
,118
11,769
14,863
6,407

df
1
1
1
1
1
1
1
1
1
1
1
1
26
26
26

Mean
Square
2,609
,532
,202
,504
1,437
2,133E-02
1,680E-02
1,232
,346
,683
1,782
,118
,453
,572
,246

F
5,764
,931
,820
1,114
2,513
,087
,037
2,155
1,403
1,508
3,118
,478

Sig.
,024
,343
,373
,301
,125
,771
,849
,154
,247
,230
,089
,496

ANOVA medidas repetidas


Between-Subjects Factors

Sexo

1
2

Value
Label
Hombre
Mujer

N
12
18

Tests of Between-Subjects Effects


Measure: MEASURE_1
Transformed Variable: Average

Source
Intercept
EDAD
T_EVOLUC
SEXO
Error

Type III
Sum of
Squares
1139,147
55,247
11,519
,346
131,876

df
1
1
1
1
26

Mean
Square
1139,147
55,247
11,519
,346
5,072

F
224,589
10,892
2,271
,068

Sig.
,000
,003
,144
,796