I have attended different universities during my study programme and for different conference and seminars, where I notice a critical number always suggested by almost every scientist/professors i.e. 12 error degree of freedom. Why this error degree of freedom should always be greater than 12 and what if it is 8 or 10 ? Can anyone start his/her research work taking Randomized Block Design with 5 treatments and 3 blocks; if yes then why the number 12 is important and if no then why?
Hello Jagadish,
In brief, more df for the error term means a more powerful test of the respective effect (when using F-ratios). I can't say why 12 might be considered by some to be a magical number. I can say that, more is better (within reason, of course).
For "classic" RB design (e.g., no replications per combination of "A" and "B" factors), since you do not have within-cell replications, the only way to increase error df for the respective main effect tests is to have more levels of the "A" and "B" factors.
On the other hand, if you have replications within cells, then the number of levels of the two factors becomes less important as concerns the potential for statistical power.
Good luck with your work.
Sir you mean there is no such limit that error df must be more than 12! Thank you sir for your valid explanation David Morse
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology