If there are at least 4-5 factors to consider, there will be too many samples. I have read something about 2k factorial level design, and some researchers used to screen out the important factors.
IF you have 4-5 parameter then choose three important from it and go with BOX-Behnken design method.
4-5 parameters or varaiables?, Variables/Factors are choosed not parameters.
Variable and parameter are two terms. A variable is an entity that changes with respect to another entity. A parameter is an entity which is used to connect variables.
Difference between variables and parameters. A variable represents a model state, and may change during simulation. A parameter is commonly used to describe objects statically
A variable is a real world value with a measureable quantity whereas a parameter is an entity that we may or may not be able to measure.
4-5 parameters or varaiables?, Variables/Factors are choosed not parameters.
Variable and parameter are two terms. A variable is an entity that changes with respect to another entity. A parameter is an entity which is used to connect variables.
Difference between variables and parameters. A variable represents a model state, and may change during simulation. A parameter is commonly used to describe objects statically
A variable is a real world value with a measureable quantity whereas a parameter is an entity that we may or may not be able to measure.
4-5 parameters or varaiables?, Variables/Factors are choosed not parameters.
Variable and parameter are two terms. A variable is an entity that changes with respect to another entity. A parameter is an entity which is used to connect variables.
Difference between variables and parameters. A variable represents a model state, and may change during simulation. A parameter is commonly used to describe objects statically
A variable is a real world value with a measureable quantity whereas a parameter is an entity that we may or may not be able to measure.
Prescreening and optimization of the most effective independent variables of operating conditions of any process using conventional methods of unplanned approaches that involve varying one variable at a time and remaining the others constant is expensive and time consuming, specifically when a huge number of parameters are to be examined and evaluated. Therefore experimental design (DOE) using central coposite design (CCD) combined with response surface methodology approach is an efficient procedure for planning experiments so that data obtained can be analyzed to yield valid and objective conclusions. DOE begins with determine the objectives of an experiment and selecting the process factors for the study. An experimental design is the layout of a detailed experimental plan in advance of doing the experiment.
All input factors are set at two levels each. These levels are called high and low or “+1” and “-1, respectively. there are two types of expermental design; The full factorial design and fractional design of experiments (DOE). A design with all possible high/low combinations of all input factors is called a full factorial design in two levels. If there are k factors, a full factorial design has 2k runs. Table 1 present the results for number of factors from 2 to 7.
Table 1: Number of Runs for 2k Full Factorial
Number of Factors
Number Of Runs
2 4
3 8
4 16
5 32
6 64
7 128
When the number of factors is 5 or greater, a full factorial design requires a large number of runs and is not very efficient. However A huge number of experiments should be performed. For example if you have 10 factor designs at two levels, means 1024 experiments are necessary to run. However these experiments don’t always result in useful information. Consequently, they are wasteful of time and resources mainly in the screening stage. Therefore fractional factorial design or a placket-Burman design is a better choice for 3 or more factors.
Table 1: Number of Runs for 2k Full Factorial
Number of Factors Number Of Runs
2 4
3 8
4 16
5 32
6 64
7 128
Hey guys, thank for your feedback ^^.
Actually I have yet to determine my parameters. Not so sure about the terms, like temperature, concentration, time. etc etc. I do have little experience in response surface methodology RSM when I was doing my final year degree. That things came when I am going to do optimisation. If optimisation came together, then it will be great. However, my knowledge about RSM is not that depth yet. I have encountered effects which is compromising each other, that had made my reliability < 0.80 during optimisation process. Will it be a problem?
If you are looking for screening factors, the choice is fractional 2-level factorial designs with the aedequate resolution to discriminate clearly main effects from two factor interactions, this will reduce the number of experiments you need to run. Once you have 2-3 factors form the screening, you can expand the already made experiment (the fractional factorial) to a RSM with some additional experiments.
You can go further. If you have any prior information about the 'true' model, maybe from literature or experience, you can use Optimal Design of Experiments to create the best design to estimate the effects of interests (main or interactions). In addition, you can also create the best design to make predictions of new data if that is your case.
By using Optimal Design of Experiments you can set the sample size you want/can use (according to your budget, for example), and construct the best design for your situation.
There are optimal designs for response surface analysis, split-plot designs, and screening designs.
Hi Alan. this is something new for me and I need further reading about this. Any suggestion for further reading?
This book is an excellent starting point:
"Optimal Design of Experiments: A case study approach" by Goos and Jones.
Each chapter is like a detective novel. It tells you a complete story, from the beginning of the problem to its solution. Then, they explain you the mathematical concepts in a ver friendly way. And finally, they provide you with the algorithms to construct the designs. You can create all the optimal designs that appear in the book by using JMP.
hey study the chapter optimization of pharmaceutical dosage form in Moder Pharmaceutics, Banker, Dekker publications.
factorial designs will help u to resolve the issue. u find out which factor is highly infulenital on the product quality which is then supported by ANOVA, F test, tukey test will often followed by the researchers. moreover factorial designs best dealt in pharmaceutical statistics by stanford bolton, dekker publications.
If you want to test 4-5 variables/factors at the same time, you can use a Res-5 5 factor fractional factorial design. This will give you knowledge about all main effects and every 2-way interaction. This is a linear model only.
If you want to test for curvature in your design, it will require you to use an optimal response surface. This will require a minimum of 21 runs. A 26 run design will be much better.
I do not believe researchers can screen out the most important factors that influence processes. It makes no sense. A good foundational knowledge of what to do, when to do, why do and how to do is strongly recommended. You can do a good or decent screening with the wealth of information you have acquired about the nature of research you embark on.
You may want to consider Plackett–Burman method: https://en.wikipedia.org/wiki/Plackett%E2%80%93Burman_design https://www.itl.nist.gov/div898/handbook/pri/section3/pri335.htm
- Badges
- Science topic