Analysis of DOE Results and Model Finalization

v9991

Trusted Information Resource
How to resolve following situation?
1. i find the model-significant; lack of fit-non significant; but curvature-significant.
problem, is that how to assess the curvature. (design already has enough center points); should we go for CCD?

the results are pretty much fine...is there a way to build argument to conclude the model without exploring the curvature part?

2. i find 4 factors of 'equal effect', but their quantum of effect is not significant; You can imagine them to be a tower one of over other on the half normal plot...how to handle them....
 

Miner

Forum Moderator
Leader
Admin
  1. The logical next step would be to augment your experiment by adding axial points to create a CCD, either a circumscribed (preferred) or face centered (if levels are restricted) design. You should add additional centerpoints because this will create a second block, and you will need the center points to detect a potential block effect.
  2. Not sure what you mean by quantum of effect. Can you attach the ANOVA table and half normal plot?
 

v9991

Trusted Information Resource
  1. The logical next step would be to augment your experiment by adding axial points to create a CCD, either a circumscribed (preferred) or face centered (if levels are restricted) design. You should add additional centerpoints because this will create a second block, and you will need the center points to detect a potential block effect.
  2. Not sure what you mean by quantum of effect. Can you attach the ANOVA table and half normal plot?
1. is there any way to conclude without extending the CCD or resolving the curvature part of the model.
2. am enclosing the half normal plot.
Analysis of variance table [Partial sum of squares - Type III]
Sum of Mean F p-value
Source Squares df Square Value Prob > F
Model 52.00 2 26.00 57.20 < 0.0001 significant
A- 50.00 1 50.00 110.00 < 0.0001
B- 2.00 1 2.00 4.40 0.0692
Residual 3.64 8 0.45
Lack of Fit 2.97 6 0.49 1.48 0.4553 not significant
Pure Error 0.67 2 0.33
Cor Total 55.64 10
 

Attachments

  • half-normal plot.jpg
    half-normal plot.jpg
    23.8 KB · Views: 182
Top Bottom