## Braveheart! [R for BE/BA]

Hi all,

The solution I posted is a total winner. Where do I collect my medal?

Will be happy to learn of performance on other datasets, in particular of differences between results or non-convergence (which is often something related to instability caused by the initial guess being too far from the optimum). Obviously, with this implementation the input format has to meet the expectation of the algo. For example, T and R coded as T and R, not A and B, in this implementation. All this is unrelated to the task of this thread and can be rather easily worked out as well.

A slight change this morning, for the initial guesses, less clumsy and more likely (in the nonrestricted fashion ) to reflect the optimal solution:

 Some.Initial.Guesses=function(foo) {  #guess varT  D1=subset(D, (D$Trt=="T")) m=lm(log(Y)~factor(Seq)+factor(Per) , data=D1) varT=anova(m)["Residuals", "Mean Sq"] #guess VarWR D1=subset(D, (D$Trt=="R"))  m=lm(log(Y)~factor(Seq)+factor(Per)+factor(Subj)  , data=D1)  varWR=anova(m)["Residuals", "Mean Sq"]   #guess varBR  D1=subset(D, (D\$Trt=="R"))  m=lm(log(Y)~factor(Seq)+factor(Per)   , data=D1)  varBR=anova(m)["Residuals", "Mean Sq"] - varWR  #guess varRT (between), must tbe closer to the other two betweens.  ##perhaps mean is a decent guess.  varRT=0.5*(varT+varBR)  rslt=c(varT, varBR, varWR, varRT)  return(rslt) } 

I could be wrong, but...

Best regards,
ElMaestro

"Pass or fail" (D. Potvin et al., 2008)