Re: Design of Experiments: Lack of fit
- From: "Arto Huttunen" <artohuttunen@xxxxxxxxxxx>
- Date: Tue, 3 Jul 2007 17:29:52 +0300
"Danielle D." <ventikas@xxxxxx> wrote in message
news:28478713.1183452049317.JavaMail.jakarta@xxxxxxxxxxxxxxxxxxxxxxxxx
Hi,
I have a simple question about design of experiments. I made 20
experiments, model was 2^4 (2-level factorial) + 4 experiments at the
middle of the model. So I have now the linear models for output. For data
analysis I use program Design Expert. The program indicates that the "Lack
of Fit F-value" implies the Lack of Fit is significant.
Does it mean that the 2-level experimental design does not catch all the
details of the phenomena and I need more levels?
Are the models totally wrong and useless? Or perhaps is the significance
of different parameters still similar, only the slightly different? And
the most important question, if I can't add more levels, is my
experimental study totally useless?
Thanks in advance
Danielle D.
The purpose of DOE is to extract the significant variables out of
a set of experiments.
It can be done normally with two level designs.
Your extra middle level experiments may cause loss of significance in linear
model.
Run your analysis data with just 16 base experiments using normal linear
regression analysis.
Then use the extra 4 to estimate nonlinear effects.
Arto(dot)Huttunen(at)estlab.com
.
- References:
- Design of Experiments: Lack of fit
- From: Danielle D.
- Design of Experiments: Lack of fit
- Prev by Date: Re: Design of Experiments: Lack of fit
- Next by Date: Re: hotelling t2
- Previous by thread: Re: Design of Experiments: Lack of fit
- Next by thread: Re: Design of Experiments: Lack of fit
- Index(es):
Relevant Pages
|
