I have analyze my data of RSM using Centeral Composite Design (CCD). After applying ANOVA my results are significant but "Lake of fitness" is also showing significant. So can anyone explain me how I can adjust it? I shall be very thankful.
You need to change the functional form of the mean model and reanalyze, one simplest strategy is to do a Box and Cox transformation of the dependant variable and see if this altered postulated model leads to null hypothesis being retained in LOF F test, otherwise you need to alter the functional form of the independent variables , e.g replace X by ln X etc. and see whether Lack of fit in the altered model is now statistically insignificant.
Dear Hamza Noushahi,
You have to see whether lack of fit in the distorted model is now statistically insignificant or not? Good strategy you can have Box and Cox transformation of the Dependant variable and must see if this distorted postulated model leads to unsound hypothesis being retained in F test or not?
Ashish
It would had been helpful if I know the software u used in analyzing your experimental data which would had given me a clue to profound solution to your problem. Anyway a good way to solve the problem is by checking the diagnostic tool to see if an outlier emarged at the outlier t plot from whence a box cos transformation can be applied. But generally, when u have a lack of fit that is significant, the problem is usually from the result obtained from the centre point of the experimental run. Try and locate the experimental runs that fall under the centre point since u used CCD. The result obtained is either above the maximum values that should be obtained for such runs or below the minimum values that should be obtained for the experimental runs. The experiment based on the centre point needs to be rerun and analyze again.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA