# not a function file
numPart=linspace(1,10,10);
numTreat1=ones(10,1)*1;
numTreat2=ones(10,1)*2;
numTreat3=ones(10,1)*3;
parts=floor(rand(10,1)*3)/100;
treat1=0.02*ones(10,1);
treat2=0.01*ones(10,1);
treat3=-0.03*ones(10,1);
errors=floor(rand(60,1)*3)/100;

numParts=[numPart, numPart, numPart, numPart, numPart, numPart]';
numTreats=[numTreat1; numTreat1; numTreat2; numTreat2; numTreat3; numTreat3];

parts=[parts;parts;parts;parts;parts;parts];
treats=[treat1; treat1; treat2; treat2; treat3; treat3];
value=parts+treats+errors+10.0e0;

anovan(value, [numParts, numTreats])

My goal is to calculate 2-way anova tests for Gauge R&R study. In my experience, anovan always returns Fval=0.000 and p-value NaN, for every input value. This behaviour was also found when inputting dataset copied from textbooks or literature. I have written some lines to provide an example without copying copyrighted material. All the datasets I can generate with the lines above were checked by me against results calculated with R software, and in every case I am getting a result different from Fval=0.000.