Hi SEs,
I created a Journal on Design Diagnostics I want to share with you. It is also posted to be available in the community: MJMP_DesignDiagnostics_2015_10_09 09Okt2015 12_03_44.zip
May this is useful for others as it has different comparisons you could use also onsite and enlarge if you want to dive in deeper than it is possible in 45 minutes.
Regards,
Martin
Here is the description of the Content:
This Journal has been presented in my German Mastering JMP Webinar on Design Diagnostics on 09.10.2015. The Journal is in English, so everyone can benefit from.
I hope I added enough description so everyone can get used to it. Within the journal there are further linked resources on Webinars, Blogs, Books, Papers and Documentation. There are also tables within the zip file, which are listed here:
- MJMP_DesignDiagnostics_2015_10_09.jrn
- Widget Factors.jmp
- Custom Design_30run.jmp
- DSDlikeDesign_30run.jmp
- PBlikeAugmentedDesign_30run.jmp
- Custom_D_Optimal.jmp
- Custom_I_Optimal.jmp
- Relative Std Error of Estimate.jmp
In the Journal I will not talk about Clinical Trial Designs, Designs for non-linear or generalized linear models, pure mixture designs, or covering Arrays. However there are links to a blog entry on DOE for non-linear models as well as alink to MJMP webcat on Mixture Designs. The Content is as follows:
- Some General remarks on Fisher's design principles, as well on what effect the probability to detect a significant factor, that is when a change in the Level of this Parameter will cause a practical important Change in the Response. This probability is the power, you learn on what it is based and how it could be increased.
- Compare Designs in there ability to estimate model effects. Especially to Highlight that not all designs can estimate all model effects like two-factor interactions or quadratic effects. Then a comparison on power Shows that you should make sure to compare apples with apples, that is to compare designs with same number of runs and same model effects
- As comparison we look into different Kind of designs. Custom design, Placket-Burman-like design and Definitive Screening like design with each 30 rund (Don McCormack's English Webcast on Design Diagnostics will Show you exactly how to generate those designs:
(http://www.jmp.com/en_gb/events/ondemand/mastering-jmp/design-diagnostics.html)) - The Focus is on Evaluate Design Platform, what can be modelled and Power Analysis, how good it can be estimated
- The next comarison will talk about D-Optimal and Alias-Optimal designs on the importance of taking care of aliasing. This part is based on a blog from Ryan Lekivetz
(http://blogs.sas.com/content/jmp/2014/06/12/alias-optimal-versus-d-optimal-designs/). It also describes the Definitive Screening Designs briefly. - The Focus here is on Optimality criterion, Color Map on Correlations, and Estimation Efficiency
- The last comparison example is based on a "robust and optimal process Experiment" example in the book written by Bradley Jones and Peter Goos
(http://www.jmp.com/en_gb/events/ondemand/mastering-jmp/using-definitive-screening-designs-to-get-more-information-from-fewer-trials.html) - It will Highlight i-optimal designs and their Advantage against d-optimal designs regarding reduced prediction variance.
- The Focus is on estimation Efficiency, Design Efficiency and Fraction of Design Space Plot.
