names(Y1.ls)
#here my code starts:
#some helper functions
splitchar <- function(x){split(x,x)}
class.summary.f <- function(class.o){
	lapply(splitchar(perf.metrics),function(metr)	evaluation(class.o,meas=metr)@score	) #metr is misclassification rates and so on
	}

#perfomance metric to use:
perf.metrics <- c("misclassification", "sensitivity", "specificity")
set.seed(123)
results.of.50.iteration	<- lapply(1:50,function(i){
			Y.ls=mapply(function(v1,s1){
						sample(v1,s1,replace=T) #sample  sample
					},Y1.ls,Y.size.ls)

			X <- X1[unlist(Y.ls),]	#create "simulated"(desired class sizes) subset of the original data
			Y <- Y1[unlist(Y.ls)]

			lset <-  GenerateLearningsets(y =Y , method = "bootstrap", niter=1, strat = TRUE)


			y<-Y


			### clasification code below here:

			simurealCV <-list() #more flexible data structure than vecor, you don't need 'j'(i.e count)

			####################################################
			##### t-test as preliminary variable selection #####
			####################################################

			selttest<-GeneSelection(X, y, learningsets = lset, method="t.test" , scheme="one-vs-all")





			### dlda t-test ###

			#j<-1, you don't need count
# 1
simurealCV <- c(simurealCV,list('dldaCMA_NG20' 	=class.summary.f(try(
	#you don't have to use name for classiffier,  list(class.summary.f(....  will work
						classification(X, y, learningsets = lset,
		genesel=selttest, classifier = dldaCMA, nbgene = 20)
)
)))


#next classifier
simurealCV <- c(simurealCV,list('dldaCMA_NG50'=class.summary.f(try(
						classification(X, y, learningsets = lset,
								genesel=selttest, classifier = dldaCMA, nbgene = 50)
				)
)))

simurealCV
		})


results.of.50.iteration.df <- melt(results.of.50.iteration)
head(results.of.50.iteration.df,10)
#value                L3           L2 L1
#1  0.00000000 misclassification dldaCMA_NG20  1
#2  1.00000000       sensitivity dldaCMA_NG20  1
#3  1.00000000       specificity dldaCMA_NG20  1
#4  0.02941176 misclassification dldaCMA_NG50  1
#5  1.00000000       sensitivity dldaCMA_NG50  1
#6  0.66666667       specificity dldaCMA_NG50  1
#7  0.10810811 misclassification dldaCMA_NG20  2
#8  1.00000000       sensitivity dldaCMA_NG20  2
#9  0.20000000       specificity dldaCMA_NG20  2
#10 0.10810811 misclassification dldaCMA_NG50  2

# one value for L3+L2+L1 combination because only 1 cross validation iteration
#this frame can be aggerageted futher
#to average cross sampling iterations, i.e. get a  data frame with columns for misclassification mean(M),SD, sensitivity and specifity
#by classifier
results.of.50.iteration.df.M <-dcast(results.of.50.iteration.df,L2~L3,fun=mean,na.rm=T) $na.rm remove NA i.e. failed classifier runs
head(results.of.50.iteration.df.M)
#L2 misclassification sensitivity specificity
#1 dldaCMA_NG20        0.07225663   0.9790976   0.4363333
#2 dldaCMA_NG50        0.07926080   0.9657780   0.4970000

#to calculate standart deviation cross sampling iterations
dcast(results.of.50.iteration.df,L2~L3,fun=sd,na.rm=T)
#L2 misclassification sensitivity specificity
#1 dldaCMA_NG20        0.04422964  0.02726753   0.3307672
#2 dldaCMA_NG50        0.04619228  0.03935401   0.3115991

#another more flexible way to get varios aggregation cross sampling iterations
results.of.50.iteration.df.aggr <- melt(
ddply(results.of.50.iteration.df,.(L2,L3),function(x) c(mean=mean(x$value,na.rm=T),sd=sd(x$value,na.rm=T))),
meas=c('mean','sd'))
#results.of.50.iteration.df.aggr
#L2                L3 variable      value
#1  dldaCMA_NG20 misclassification     mean 0.07225663
#2  dldaCMA_NG20       sensitivity     mean 0.97909763
#3  dldaCMA_NG20       specificity     mean 0.43633333
#4  dldaCMA_NG50 misclassification     mean 0.07926080
#5  dldaCMA_NG50       sensitivity     mean 0.96577800
#6  dldaCMA_NG50       specificity     mean 0.49700000
#7  dldaCMA_NG20 misclassification       sd 0.04422964
#8  dldaCMA_NG20       sensitivity       sd 0.02726753
#9  dldaCMA_NG20       specificity       sd 0.33076716
#10 dldaCMA_NG50 misclassification       sd 0.04619228
#11 dldaCMA_NG50       sensitivity       sd 0.03935401
#12 dldaCMA_NG50       specificity       sd 0.31159909


#to get table with means for the all perfomance metrics, and standart deviation for misclassifcation rate
results.of.50.iteration <- dcast(subset(results.of.50.iteration.df.aggr,variable=='mean'|(variable=='sd'&L3=='misclassification')),
		L2~L3+variable,fun=mean)
results.of.50.iteration
#L2 misclassification_mean misclassification_sd sensitivity_mean
#1 dldaCMA_NG20             0.07225663           0.04422964        0.9790976
#2 dldaCMA_NG50             0.07926080           0.04619228        0.9657780
#specificity_mean
#1        0.4363333
#2        0.4970000



write.csv(results.of.50.iteration, na="", #NA will be empty cells
		file = "E:/bioconductor/results.csv")