Sie sind auf Seite 1von 12

Stat 3520: Design of Experiments-1

Analysis of Fixed Eect Model

Oct 7, 2011

Outline

ANOVA - An Example R Code

Example
A product developer is investigating the tensile strength of a new synthetic ber that will be used to make cloth for mens shirts. Strength is usually aected by the percentage of cotton used in the blend of materials for the ber. The engineer conducts a completely randomized experiment with ve levels of cotton content and replicated the experiment ve times. The data are shown in the following table. Cotton Wt. % Rep1 Rep2 Rep3 Rep 4 Rep5 15 7 7 15 11 9 20 12 17 12 18 18 25 14 19 19 18 18 30 19 25 22 19 23 35 7 10 11 15 11 Is there evidence to support the claim that cotton content aects the mean tensile strength? Use = 0.05. Perform residual analysis.

Questions
Problem? Factors and levels? Treatments? How the experimental principles are employed? Response? Hypothesis? Exploratory Analysis? Type of Analysis? Expected interpretation of the results? Need to check any assumptions? What further analysis could be done?

Simple Analysis
Cotton Wt% Mean Variance 15 9.8 11.2 20 15.4 9.8 25 17.6 4.3 30 21.6 6.8 35 10.8 8.2

Tensile Strength

10

15

20

25

15

20

25 Cotton Wt

30

35

Sum of Squares

Correction Factor, CF= Total SS = 72 + 72 +

(7+7+...+15+11)2 = 3762 /25 = 55 . . . + 152 + 112 CF = 636.96

5655.04

Treatment sums: 49, 77, 88, 108, 54 Treatment SS = (492 + . . . + 542 )/5 CF = 475.76 Error SS = Total SS - Treatment SS = 161.20

ANOVA

Source d.f SS MS Fc Treatments 4 475.76 118.94 14.76 Error 20 161.20 8.06 Total 24 475.76 Prob(Fc > F ) = 9.128e 06) (alternatively look for F0.05,4,20 = 2.87, F0.01,4,20 = 4.43 Treatments are highly signicant

Normal Probability Plot of Residuals

Normal Q-Q Plot

Sample residual quantiles

-4
-2

-2

-1

0 Theoretical Quantiles

Plot of Residual over Time

Residuals

-4

-2

10

15

20

25

Time Order

Plot of Residual over Fitted Values

Residuals

-4
10

-2

12

14

16

18

20

22

Fitted Values

Practical Interpretation of Results

Need to do use more formal techniques to get better interpretation Factors involved can be qualitative or quantitative Experimenter interested to predict the response for an intermediate level of the factor in case of quantitative factor. A scatter diagram will give some idea about the relationship Fit linear regression model to data and see how it works If needed you may use to t the data such as y = 0 + 1 x + 2 x 2

Fitting linear and quadratic model to the data


Linear Model: y = 10.940 + 0.164x Quadratic Model: y = 39.98857 + 4.59257x 0.08857x 2

Tensile Strength

10
15

15

20

25

20

25 Cotton Wt

30

35

Das könnte Ihnen auch gefallen