F vales come out as Infinity using the anovan function.

I'm trying to use the anovan function to try and get some statistical data on a partial factorial DOE. However, the F values come out as infinite. I was wondering what that could mean, and what the typical causes are. Would this be a problem with the design of experiments itself, the code, or does it have some sort of statistical significance?

Answer 1

[dpb]

You've tried to estimate 15 terms with only 15 data points; note you have 0 DOF left for the error term. Hence, the denominator in the F statistic is 0-->inf for the estimate.

The problem is in at least the analysis of trying to estimate more effects than what you have data for which to do; the issue could be you simply mis-specified those terms in the model that weren't intended to be estimated or it could be even more fundamental that the design points are such that you can't effectively estimate much of anything; we can't tell, we don't have the design available.

The normal procedure is to prepare a null hypothesis to test including which factors and interactions and then design, execute, and analyze appropriately to test that null hypothesis. Far too often, however, people simply take a bunch of data and then try to force it into the model; we've no way of knowing which took place here (or something in between).