Good Day,this is urgent please, i want to use statgraphics trial version to study a part of my work. I am considering different stratified layers in a particular oil reservoir, i considered different factors that affects recovery and i got down to three important parameters for each layer. This brings me to a total of 9 continuous factors because the factors would have to be run together to choose the optimal combinations of runs for recovery. two of these factors from previous literature have non-linear response. Please how do i model it. I chose multi-level factorial and wanted to give the two factors 5-levels while leaving the other seven factors at 3 levels, the runs were too much. I was planning on using D or I-Optimal designs Please help! This is my first time at trying DOE. Thank you. Please other soft-wares are welcome!
Dear Mary,
it's not that easy to explain the whole theory to you in DoE in a short time. Besides, I don't think you have the time for it. So I try to limit myself to a few essential things. First of all you have to think about the scatter of your target quantity. Depending on the scatter, you can then use software to determine the sample size required to detect significant effects (see power analysis). Furthermore, I wonder if the nine factors you mentioned are actually statistically significant. In other words: Usually you make screening experiments to answer exactly this question. It would be nice if you could limit yourself to a maximum of 5 factors. With these factors you plan a "new" experimental design for modeling. Here you can either use central composite designs, or if you cannot perform certain experiments, you can also use D-optimal experimental designs. The advantages of D-optimal experimental designs are that you can choose the tests by yourself under certain conditions. You can also choose the distances between the factor levels. So D-optimal experimental designs are quite flexible.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running