i want make screening of variables by Taguchi and then use Box-Behnken design for the significant variables. is the elimination of variables is based on main plot or S/N ratio or ANOVA of S/N ratio or ANOVA of response. And how i get the optimum condition from RSM.
Hi Hadeel,
The test of significance of your variables should be made based on ANOVA for your response. Actually ANOVA for S/N is the same as ANOVA for your response. All about your collected data (response). I recommend you use fractional factorial design instead of using Taguchi then use Box-Behnken design. Why I am saying that because fractional factorial design is used to estimate the individual effect as well as interaction effect. For example, if you have three variables (A, B, C) and variable (A) has no significant effect on the response as an individual factor but it has a significant effect in interaction with variable (B). In this case you should include the variable (A) in Box-Behnken design to obtain a good regression model. If you do not include variable (A) in Box-Behnken design then the lack-of-fit test will be significant. Thus there is a need to re-conduct your experiment with including variable A. This leads to loss time, money and effort.
Kind regards,
Omar
Hi Hadeel,
The test of significance of your variables should be made based on ANOVA for your response. Actually ANOVA for S/N is the same as ANOVA for your response. All about your collected data (response). I recommend you use fractional factorial design instead of using Taguchi then use Box-Behnken design. Why I am saying that because fractional factorial design is used to estimate the individual effect as well as interaction effect. For example, if you have three variables (A, B, C) and variable (A) has no significant effect on the response as an individual factor but it has a significant effect in interaction with variable (B). In this case you should include the variable (A) in Box-Behnken design to obtain a good regression model. If you do not include variable (A) in Box-Behnken design then the lack-of-fit test will be significant. Thus there is a need to re-conduct your experiment with including variable A. This leads to loss time, money and effort.
Kind regards,
Omar
You can obtain the optimum conditions by using desirability function after building your regression model.
thanks for answering Dr. Omar
Actually in the beginning i decided to use Box-Behnken, when i do the regression for the response i get the ANOVA too and there is the test for liner, interaction and quadratic terms if you that what you mean is that correct ?. you see am still beginner in MIINTAB :D
Hi Hadeel
I agree with @Omar Ahmed Mohamed.
Many a times mistake is committed by researchers while performing screening using Taguchi method where they assign the factors directly to the columns of an orthogonal array without bothering whether it is confounded with the effect of any other factor column which could wrong interpretation of results. You can use Taguchi method if proper care is taken in designing the experiments. In the Taguchi analysis, base is raw data so whether it is ANOVA of means or ANOVA of S/N ratio, it is same.
Box-Behnken design is prominently used to eliminate many insignificant factors.
- Badges
- Science topic
- Similar topics
- Chemistry
- Nanotechnology