Effects that cannot be estimated separately from one another are said to be confounded. Confounding occurs when you use a fractional factorial design, because you do not run all factor level combinations. For example, if factor A is confounded with the 3-way interaction BCD, then the estimated effect for A is the sum of the effect of A and the effect of BCD. These effects are also said to be aliased. The alias structure describes the confounding that occurs in the design.
Effects that are aliased, or confounded, cannot be estimated separately from one another. For example, if the two-way interaction, BC, is confounded with the three-way interaction, ADE, you will not be able to tell whether a significant effect is due to the BC interaction or the ADE interaction.
The key to the alias structure is the identity statement, for example, I + ABCDE. To determine which effects are confounded, multiply the term of interest by the identity statement and then eliminate the squared terms. For example, to find the term that BC is confounded with:
(BC)(I ABCDE) = BC AB2C2DE = BC ADE
Therefore, BC and ADE are confounded with one another.
Example of creating a blocked design
main topic interpreting results session command see also
You would like to study the effects of five input variables on the impurity of a vaccine. Each batch only contains enough raw material to manufacture four tubes of the vaccine. To remove the effects due to differences in the four batches of raw material, you decide to perform the experiment in four blocks. To determine the experimental conditions that will be used for each run, you create a 5-factor , 16-run design, in 4 blocks .
1 Choose Stat >DOE > Factorial > Create Factorial Design.
2 From Number of factors, choose 5.
3 Click Designs.
4 In the box at the top, highlight the line for 1/2 fraction.
5 From Number of blocks, choose 4. Click OK.
6 Click Results. Choose Summary table, alias table, design table, defining relation. Click OK in each dialog box.
Session window output
Fractional Factorial Design
Factors: 5 Base Design: 5, 16 Resolution with blocks: III
Runs: 16 Replicates: 1 Fraction: 1/2
Blocks: 4 Center pts (total): 0
* NOTE * Blocks are confounded with two-way interactions.
Design Generators: E = ABCD
Block Generators: AB, AC
Defining Relation: I = ABCDE
Alias Structure
I + ABCDE
Blk1 = AB + CDE
Blk2 = AC + BDE
Blk3 = BC + ADE
A + BCDE
B + ACDE
C + ABDE
D + ABCE
E + ABCD
AD + BCE
AE + BCD
BD + ACE
BE + ACD
CD + ABE
CE + ABD
DE + ABC
Design Table (randomized)
Run Block A B C D E
1 1 + - - - -
2 1 - + + - +
3 1 - + + + -
4 1 + - - + +
5 3 - + - + +
6 3 + - + - +
7 3 + - + + -
8 3 - + - - -
9 4 + + + + +
10 4 + + + - -
11 4 - - - + -
12 4 - - - - +
13 2 + + - - +
14 2 - - + + +
15 2 - - + - -
16 2 + + - + -
|
Interpreting the results
The first table gives a summary of the design: the total number of factors , runs , blocks , replicates , center points , and resolution . After blocking, this is a resolution III design because blocks are confounded with 2-way interactions.
Because you chose to display the summary and design tables, Minitab shows the experimental conditions or settings for each of the factors for the design points. When you perform the experiment, use the order that is shown to determine the conditions for each run.
The first four runs of your experiment would all be performed using raw material from the same batch (Block 1). For the first run in block one, you would set Factor A high, Factor B low, Factor C low, Factor D low, and Factor E low, and measure the impurity of the vaccine.
Minitab randomizes the design by default, so if you try to replicate this example your run order may not match the order shown.
Design resolution
Describes the extent to which effects in a fractional factorial design are aliased with other effects. When you run a fractional factorial design, one or more of the effects are confounded, meaning they cannot be estimated separately from one another. In general, you want to use a fractional factorial design with the highest possible resolution for the amount of fractionation required. For example, it is usually better to choose a design where main effects are confounded with 3-way interactions (Resolution IV) over a design where main effects are confounded with 2-way interactions (Resolution III).
Resolution III, IV, and V designs are the most common:
Resolution III
|
No main effects are aliased with any other main effect, but main effects are aliased with 2-factor interactions.
|
Resolution IV
|
No main effects are aliased with any other main effect or 2-factor interactions, but some 2-factor interactions are aliased with other 2-factor interactions and main effects are aliased with 3-factor interactions.
|
Resolution V
|
No main effects or 2-factor interactions are aliased with any other main effect or 2-factor interactions, but 2-factor interactions are aliased with 3-factor interactions and main effects are aliased with 4-factor interactions.
|
Example of creating a Plackett-Burman design with center points
main topic interpreting results session command see also
Suppose you want to study the effects of 9 factors using only 12 runs , with 3 center points . In this 12 run design, each main effect is partially confounded with more than one 2-way interaction.
1 Choose Stat > DOE > Factorial > Create Factorial Design.
2 Choose Plackett-Burman design.
3 From Number of factors, choose 9.
4 Click Designs.
5 From Number of runs, choose 12.
6 In Number of center points per replicate, enter 3.
7 Click Results. Choose Summary table and design table. Click OK in each dialog box.
Session window output
Plackett - Burman Design
Factors: 9 Replicates: 1
Base runs: 15 Total runs: 15
Base blocks: 1 Total blocks: 1
Center points: 3
Design Table (randomized)
Run Blk A B C D E F G H J
1 1 - - - + + + - + +
2 1 + + + - + + - + -
3 1 + - + - - - + + +
4 1 + - + + - + - - -
5 1 - + + - + - - - +
6 1 + + - + - - - + +
7 1 0 0 0 0 0 0 0 0 0
8 1 - - - - - - - - -
9 1 + - - - + + + - +
10 1 0 0 0 0 0 0 0 0 0
11 1 - + - - - + + + -
12 1 - - + + + - + + -
13 1 - + + + - + + - +
14 1 0 0 0 0 0 0 0 0 0
15 1 + + - + + - + - -
|
Interpreting the results
In the first table, Total runs shows the total number of runs including any runs created by replicates and center points . For this example, you specified 12 runs and added 3 runs for center points, for a total of 15.
Minitab does not display an alias tables for this 12 run design because each main effect is partially confounded with more than one 2-way interaction.
Minitab shows the experimental conditions or settings for each of the factors for the design points. When you perform the experiment, use the order that is shown to determine the conditions for each run. For example, in the first run of your experiment, you would set Factor A low, Factor B low, Factor C low, Factor D high, Factor E high, Factor F high, Factor G low, Factor H high, and Factor J high.
Minitab randomizes the design by default, so if you try to replicate this example your runs may not match the order shown.
Thí dụ mẫu nhân tố đầy đủ với lập lại và khối
Một nghiên cứu nhằm mục tiêu khảo sát ảnh hưởng của 3 yếu tố đến kết quả của một phản ứng hóa học : thời gian (20 , 50 phút) , nhiệt độ (1500 , 2000) và chất xúc tác (A , B) . Chỉ có đủ hóa chất để thực hiện 16 thí nghiệm (16 runs) và mỗi ngày chỉ tiến hành được 8 thí nghiệm. Do đó cần tạo một mẫu nhân tố đầy đủ với 2 lập lại và 2 khối.
Bước 1 :
1 Chọn Stat > DOE > Factorial > Create Factorial Design.
2 From Number of factors, choose 3.
3 Click Designs.
4 In the box at the top, highlight the line for ful factorial..
Number of center points per block : 1
Number of replicates for corner points : 2
Number of blocks : 2
Click OK
5 Click factors
Factor
|
Name
|
Type
|
Low
|
High
|
A
|
Time
|
Numeric
|
20
|
50
|
B
|
Temperature
|
Numeric
|
150
|
200
|
C
|
Catalyst
|
Text
|
A
|
B
|
Click OK
5 Click Results. Chọn Summary table, alias table, design table, defining relation.
6 Click OK ở mỗi hộp thoại.
Full Factorial Design
Factors: 3 Base Design: 3, 8
Runs: 16 Replicates: 2
Blocks: 2 Center pts (total): 0
Block Generators: replicates
All terms are free from aliasing.
Design Table (randomized)
Run Block A B C
1 1 + - +
2 1 - - -
3 1 + + -
4 1 - + +
5 1 - + -
6 1 - - +
7 1 + - -
8 1 + + +
9 2 + + -
10 2 + - -
11 2 + + +
12 2 + - +
13 2 - - +
14 2 - + +
15 2 - + -
16 2 - - -
Lưu kết quả mẫu xuất hiện ở worksheet
Bước 2 :
Thực hiện 16 thí nghiệm (16 runs) theo bảng mẫu ở bước 1
Sau đó nhập tiếp kết quả thực nghiệm vào worksheet đã lưu ở bước 1(worksheet YIELD.MTB)
1 Chọn Stat > DOE > Factorial > Analyze Factorial Design.
3 In Responses, enter Yield.
4 Click Graphs. Under Effects Plots, check Normal, Half Normal, and Pareto. Click OK in each dialog box.
Session window output
Factorial Fit: Yield versus Block, Time, Temp, Catalyst
Estimated Effects and Coefficients for Yield (coded units)
Term Effect Coef SE Coef T P
Constant 45.5592 0.09546 477 .25 0.000
Block -0.0484 0.09546 -0.51 0.628
Time 2.9594 1.4797 0.09546 15.50 0.000
Temp 2.7632 1.3816 0.09546 14.47 0.000
Catalyst 0.1618 0.0809 0.09546 0.85 0.425
Time*Temp 0.8624 0.4312 0.09546 4.52 0.003
Time*Catalyst 0.0744 0.0372 0.09546 0.39 0.708
Temp*Catalyst -0.0867 -0.0434 0.09546 -0.45 0.663
Time*Temp*Catalyst 0.0230 0.0115 0.09546 0.12 0.907
S = 0.381847 PRESS = 5.33236
R-Sq = 98.54% R-Sq(pred) = 92.36% R-Sq(adj) = 96.87%
Analysis of Variance for Yield (coded units)
Source DF Seq SS Adj SS Adj MS F P
Blocks 1 0.0374 0.0374 0.0374 0.26 0.628
Main Effects 3 65.6780 65.6780 21.8927 150.15 0.000
2-Way Interactions 3 3.0273 3.0273 1.0091 6.92 0.017
3-Way Interactions 1 0.0021 0.0021 0.0021 0.01 0.907
Residual Error 7 1.0206 1.0206 0.1458
Total 15 69.7656
Estimated Coefficients for Yield using data in uncoded units
Term Coef
Constant 39.4786
Block -0.0483750
Time -0.102585
Temp 0.0150170
Catalyst 0.48563
Time*Temp 0.00114990
Time*Catalyst -0.0028917
Temp*Catalyst -0.00280900
Time*Temp*Catalyst 0.000030700
Alias Structure
I
Blocks =
Time
Temp
Catalyst
Time*Temp
Time*Catalyst
Temp*Catalyst
Time*Temp*Catalyst
|
Graph window output
Giải thích kết quả
Bảng phân tích phương sai cho tóm tắt về các ảnh hưởng chính và các tương tác. Nếu mô hình là trực giao (orthogonal). và không chứa hiệp biến thì Tổng bình phương liên tiếp (the sequential sums of squares (Seq SS) và tổng bình phương điều chỉnh (adjusted sums of squares (Adj SS) là bằng nhau. Xét giá trị p để xác định xem các ảnh hưởng có ý nghĩa hay không. Các ảnh hưởng được tóm tắt dưới đây :
Effect
|
P-Value
|
Significant*
|
Blocks
|
0.628
|
no
|
Main
|
0.000
|
yes
|
Two-way interactions
|
0.017
|
yes
|
Three-way interactions
|
0.907
|
no
|
* significant at alpha = 0.05
Ảnh hưởng khối không ý nghĩa cho thấy các kết quả không chĩu ảnh hưởng của ngày thu thập số liệu.
Sau khi xác định các ảnh hu7o73ngco1 ý nghĩa (ảnh hưởng chính và tương tác 2 chiều) trong bảng ANOVA, hãy chú ý bảng ước lượng các ảnh hưởng cà các hệ số. Bảng này cho thấy giá trị p gắn liền với mỗi thành phần của mô hình . Giá trị p của Time * Temp (p = 0.003), Time (p = 0.000) và Temp (p = 0.000) là có ý nghĩa . Tuy nhiên, vì cả hai ảnh hưởng chính liên quan đến tương tác , cần hiểu bản chất của sự tương tác trước khi xem xét các ảnh hưởng chính
Sai số trong bảng ANOVA có thể gồm 3 phần : (1) độ cong (curvature) , nếu có các điểm trung tâm trong số liệu, (2) sụ thiếu phù hợp (lack of fit) , nếu mô hình rút gọn ( reduced model ) phù hợp, và (3) sai số thuần (pure error), nếu có lập lại . If the residual error is just due to lack of fit, Minitab does not print this breakdown. In all other cases, it does.
Biểu đồ chuẩn , bán chuẩn và Pareto của các ảnh hưởng giúp xác định các ảnh hưởng quan trọng và so sánh mức độ tương đối của các ảnh hưởng khác nhau.
Có thể vẽ biểu đồ các sai số theo thứ tự run để kiểm tra đường khuynh hướng thời gian hay những mẫu không ngẩu nhiên khác
Chia sẻ với bạn bè của bạn: |