Verificando a Saída de Machine Learning Não Supervisionada (Cluster Analysis) Utilizando ANOVA

SAS para ANOVA

ANOVA - Análise da Variança

Para comparar mais de duas Categorias

(Para comparar duas trabalhamos com o Teste T em Excel)

Temos as categorias:

AT : Atleta

SEM: Semi-atleta

SED : Sedentário

PR: Professor

São 4 categorias, se o numero de categorias é maior que 2 temos que usar o SAS (o Excel não resolve)

Programa SAS para ANOVA

data imc_dat;

input cat $ imc corr kcal;

cards;

AT 20.2 60.7 3200

AT 21.3 54.8 3100

AT 19.3 49.6 2800

AT 21.1 52.3 3300

SEM 22.4 14.9 2600

SEM 21.9 17.8 2700

SEM 23.8 18.6 3200

SEM 24.1 15.1 3300

SED  27.3 2.5 2700

SED 23.4 4.3 2300

SED  25.2 2.3 2600

SED 26.4 2.6 3200

PR 26.2 4.1 2600

PR 24.2 2.1 2700

PR 25.4 1.9 2650

;

proc print;

run;

proc glm;

 class cat;

 model imc corr kcal  = cat;

 means cat / duncan lines;

run;

Saída do Programa SAS para ANOVA

Obs	cat	imc	corr	kcal
1	AT	20.2	60.7	3200
2	AT	21.3	54.8	3100
3	AT	19.3	49.6	2800
4	AT	21.1	52.3	3300
5	SEM	22.4	14.9	2600
6	SEM	21.9	17.8	2700
7	SEM	23.8	18.6	3200
8	SEM	24.1	15.1	3300
9	SE	27.3	2.5	2700
10	SE	23.4	4.3	2300
11	SE	25.2	2.3	2600
12	SE	26.4	2.6	3200
13	PR	26.2	4.1	2600
14	PR	24.2	2.1	2700
15	PR	25.4	1.9	2650

---

The GLM Procedure

Class Level Information
Class	Levels	Values
cat	4	AT PR SE SEM

Number of Observations Read	15
Number of Observations Used	15

---

The GLM Procedure

 

Dependent Variable: imc

Source	DF	Sum of Squares	Mean Square	F Value	Pr > F
Model	3	63.99233333	21.33077778	14.23	0.0004
Error	11	16.49166667	1.49924242
Corrected Total	14	80.48400000

R-Square	Coeff Var	Root MSE	imc Mean
0.795094	5.214802	1.224436	23.48000

Source	DF	Type I SS	Mean Square	F Value	Pr > F
cat	3	63.99233333	21.33077778	14.23	0.0004

Source	DF	Type III SS	Mean Square	F Value	Pr > F
cat	3	63.99233333	21.33077778	14.23	0.0004

---

The GLM Procedure

 

Dependent Variable: corr

Source	DF	Sum of Squares	Mean Square	F Value	Pr > F
Model	3	6829.158500	2276.386167	300.25	<.0001
Error	11	83.397500	7.581591
Corrected Total	14	6912.556000

R-Square	Coeff Var	Root MSE	corr Mean
0.987935	13.60410	2.753469	20.24000

Source	DF	Type I SS	Mean Square	F Value	Pr > F
cat	3	6829.158500	2276.386167	300.25	<.0001

Source	DF	Type III SS	Mean Square	F Value	Pr > F
cat	3	6829.158500	2276.386167	300.25	<.0001

---

The GLM Procedure

 

Dependent Variable: kcal

Source	DF	Sum of Squares	Mean Square	F Value	Pr > F
Model	3	497333.333	165777.778	1.95	0.1801
Error	11	935000.000	85000.000
Corrected Total	14	1432333.333

R-Square	Coeff Var	Root MSE	kcal Mean
0.347219	10.18210	291.5476	2863.333

Source	DF	Type I SS	Mean Square	F Value	Pr > F
cat	3	497333.3333	165777.7778	1.95	0.1801

Source	DF	Type III SS	Mean Square	F Value	Pr > F
cat	3	497333.3333	165777.7778	1.95	0.1801

---

The GLM Procedure

---

The GLM Procedure

 

Duncan's Multiple Range Test for imc

Note:This test controls the Type I comparisonwise error rate, not the experimentwise error rate.

Alpha	0.05
Error Degrees of Freedom	11
Error Mean Square	1.499242
Harmonic Mean of Cell Sizes	3.692308

Note:Cell sizes are not equal.

Number of Means	2	3	4
Critical Range	1.983	2.075	2.129

---

The GLM Procedure

 

---

The GLM Procedure

 

Duncan's Multiple Range Test for corr

Note:This test controls the Type I comparisonwise error rate, not the experimentwise error rate.

Alpha	0.05
Error Degrees of Freedom	11
Error Mean Square	7.581591
Harmonic Mean of Cell Sizes	3.692308

Note:Cell sizes are not equal.

Number of Means	2	3	4
Critical Range	4.460	4.665	4.788

---

The GLM Procedure

 

---

The GLM Procedure

 

Duncan's Multiple Range Test for kcal

Note:This test controls the Type I comparisonwise error rate, not the experimentwise error rate.

Alpha	0.05
Error Degrees of Freedom	11
Error Mean Square	85000
Harmonic Mean of Cell Sizes	3.692308

Note:Cell sizes are not equal.

Number of Means	2	3	4
Critical Range	472.3	494.0	507.0