ANOVA for design of experiments - significance between "predicted" and measured values in the domain centre

Dear all,

I'm not an expert in statistics but I need to estimate whether a difference is statistically significant. For that purpose, I know that I can make an ANOVA analysis through a Fischer test.

So… I have a multilinear model from a 2^(5-1) fractional factorial design of experiments test plan. This model is generated through the fractfact Matlab function: X=[ones(16,1),fracfact('a b c d abcd ab ac ad bc bd cd abc abd acd bcd')];

As a result, I obtained 16 operating conditions. I have 16 experiments replicated twice in a vector Y.

I have 11 replicates in the domain centre in a vector Y0

I obtained 16 coefficients from the DoE model, in a vector A, including A0, which is the value of the response in the domain centre estimated by the model.

I'm able to calculate the lack of fit error in the domain centre: Y0-repmat(A0,11,1) I'm also able to calculate the pure error on the basis of the 11 replicates in the domain centre. Now, I need to estimate whether the difference between A0 and Y0 are really significant in order to identify potential quadratic effects But I don't know which ANOVA Matlab function I should use and which arguments put inside to obtain an ANOVA table (as all those generated by the different related Matlab functions that are shown in exemples and tutorials) that will allow me to estimate the statistical difference between A0 and Y0.

Could you help me?

Thanks.