MSH

Born smart but damaged during Ph.D studies
#11
Steven is correct about why you have this issue. One workaround is to remove one term from the model. I suggest removing OF as it has the smallest Adj. MS. This frees up 1 degree of freedom for the error term allowing you to calculate p-values. Other workarounds are to either replicate the experiment, or add center points if possible.
Thanks & Regards
Miner
 

Statistical Steven

Statistician
Staff member
Super Moderator
#12
You can remove terms, but you chose the Taguchi experiment. The reason you choose this is to have a noise parameter that allows for the the signal to noise. If you wanted p-values, then you should have done a fractional factorial with center points. The L16 is a 16 run fully saturated model. A simple 2 level with center points with 5 factors can also we done with 12 runs (1/4 fraction main effects only model plus 4 center points).
 

Top