scripts/Simlation study/results-simulations-CV.R

In joakimwallmark/PolyOptimalIRT:

library(xtable)


# natMath response patterns --------------------------------------------------
U         <- scan("data/NatMath.txt","o")
N         <- length(U) # Number of examinees
Umat      <- as.integer(unlist(stringr::str_split(U,"")))
n         <- length(Umat)/N # Number of items
U         <- matrix(Umat,N,n,byrow=TRUE)
responses <- apply(U, 1, paste, collapse = "")
test <- as.data.frame(table(responses))
head(test[order(test$Freq, decreasing = T), ], 20) # most response patterns are unique

# compare bins and basis on natmath ---------------------------------------
fold <- c(5)
cy <- c(10)
meth <- c("EAP")
meth <- c("ML")
basis <- c(7, 9, 12) # 12 becomes default for natMath data
bins <- c(15, 22) # 22 becomes default for natMath data
basis <- c(15) # 12 becomes default for natMath data
bins <- c(25) # 22 becomes default for natMath data
basis <- c(12) # 12 becomes default for natMath data
bins <- c(22) # 22 becomes default for natMath data

binbasmatrix <- matrix(0, nrow = 2, ncol = 3)
binbasllmatrix <- matrix(0, nrow = 2, ncol = 3)
mirtmeanprobs <- vector("numeric", 0)
mirtlls <- vector("numeric", 0)
for (bas in 1:length(basis)) {
    for (bin in 1:length(bins)) {
        load(file = paste("data/cv/natmath ", meth,
                               " folds", fold,
                               " cyc", cy,
                               " basis", basis[bas],
                               " bins", bins[bin], ".rda", sep = ""))
        binbasmatrix[bin, bas] <- mean(res[[2]])
        binbasllmatrix[bin, bas] <- mean(res[[4]])
        mirtmeanprobs <- c(mirtmeanprobs, mean(res[[1]]))
        mirtlls <- c(mirtlls, mean(res[[3]]))
    }
}

mean(mirtmeanprobs)
mean(binbasmatrix)
binbasmatrix
binbasmatrix-mean(mirtmeanprobs)

mean(mirtlls)
mean(binbasllmatrix)
binbasllmatrix-mean(mirtlls)






# compare bins and basis on natmath ---------------------------------------
fold <- c(5)
cy <- c(10)
meth <- c("EAP")
meth <- c("ML")
basis <- c(7, 9, 12) # 12 becomes default for natMath data
bins <- c(15, 22) # 22 becomes default for natMath data
basis <- c(15) # 12 becomes default for natMath data
bins <- c(25) # 22 becomes default for natMath data

binbasmatrix <- matrix(0, nrow = 2, ncol = 3)
binbasllmatrix <- matrix(0, nrow = 2, ncol = 3)
mirtmeanprobs <- vector("numeric", 0)
mirtlls <- vector("numeric", 0)
for (bas in 1:length(basis)) {
    for (bin in 1:length(bins)) {
        load(file = paste("data/cv/natmath ", meth,
                          " folds", fold,
                          " cyc", cy,
                          " basis", basis[bas],
                          " bins", bins[bin], ".rda", sep = ""))
        binbasmatrix[bin, bas] <- mean(res[[2]])
        binbasllmatrix[bin, bas] <- mean(res[[4]])
        mirtmeanprobs <- c(mirtmeanprobs, mean(res[[1]]))
        mirtlls <- c(mirtlls, mean(res[[3]]))
    }
}

mean(mirtmeanprobs)
mean(mirtlls)
binbasmatrix-mean(mirtmeanprobs)
binbasllmatrix-mean(mirtlls)
binbasmatrix[1, 1]-mean(mirtmeanprobs)
binbasllmatrix[1, 1]-mean(mirtlls)
mean(binbasmatrix)-mean(mirtmeanprobs)
mean(binbasllmatrix)-mean(mirtlls)



# compare bins and basis on hads depr. ------------------------------------
fold <- c(10)
cy <- c(10)
meth <- c("EAP")
meth <- c("ML")
basis <- c(6, 8, 10) # 8 becomes default for natMath data
bins <- c(12, 14) # 14 becomes default for 90% of hads data
basis <- c(10, 12) # 8 becomes default for hads data
bins <- c(18, 22) # 14 becomes default for 90% of hads data

binbasmatrix <- matrix(0, nrow = 2, ncol = 3)
binbasllmatrix <- matrix(0, nrow = 2, ncol = 3)
mirtmeanprobs <- vector("numeric", 0)
mirtlls <- vector("numeric", 0)
for (bas in 1:length(basis)) {
    for (bin in 1:length(bins)) {
        load(file = paste("data/cv/hads depression ", meth,
                          " folds", fold,
                          " cyc", cy,
                          " basis", basis[bas],
                          " bins", bins[bin], ".rda", sep = ""))
        binbasmatrix[bin, bas] <- mean(res[[2]])
        binbasllmatrix[bin, bas] <- mean(res[[4]])
        mirtmeanprobs <- c(mirtmeanprobs, mean(res[[1]]))
        mirtlls <- c(mirtlls, mean(res[[3]]))
    }
}

mean(mirtmeanprobs)
mean(mirtlls)
binbasmatrix-mean(mirtmeanprobs)
binbasllmatrix-mean(mirtlls)
mean(binbasmatrix)-mean(mirtmeanprobs)
mean(binbasllmatrix)-mean(mirtlls)

mean(binbasmatrix[1:2, 1:2])-mean(mirtmeanprobs)
mean(binbasllmatrix[1:2, 1:2])-mean(mirtlls)
binbasmatrix[1:2, 1:2]-mean(mirtmeanprobs)
binbasllmatrix[1:2, 1:2]-mean(mirtlls)



# compare bins and basis on hads anxiety --------------------------------------
fold <- c(5)
cy <- c(10)
meth <- c("ML")
meth <- c("EAP")
basis <- c(6, 8, 10) # 8 becomes default for natMath data
bins <- c(12, 14) # 14 becomes default for 90% of hads data
basis <- c(10, 12) # 8 becomes default for hads data
bins <- c(18, 22) # 14 becomes default for 90% of hads data

binbasmatrix <- matrix(0, nrow = 2, ncol = 3)
binbasllmatrix <- matrix(0, nrow = 2, ncol = 3)
mirtmeanprobs <- vector("numeric", 0)
mirtlls <- vector("numeric", 0)
for (bas in 1:length(basis)) {
    for (bin in 1:length(bins)) {
        load(file = paste("data/cv/hads anxiety ", meth,
                          " folds", fold,
                          " cyc", cy,
                          " basis", basis[bas],
                          " bins", bins[bin], ".rda", sep = ""))
        binbasmatrix[bin, bas] <- mean(res[[2]])
        binbasllmatrix[bin, bas] <- mean(res[[4]])
        mirtmeanprobs <- c(mirtmeanprobs, mean(res[[1]]))
        mirtlls <- c(mirtlls, mean(res[[3]]))
    }
}

mean(mirtmeanprobs)
mean(mirtlls)
max(binbasmatrix)
max(binbasllmatrix)
binbasmatrix-mean(mirtmeanprobs)
binbasllmatrix-mean(mirtlls)
mean(binbasmatrix)-mean(mirtmeanprobs)
mean(binbasllmatrix)-mean(mirtlls)

binbasmatrix[1:2, 1:2]-mean(mirtmeanprobs)
binbasllmatrix[1:2, 1:2]-mean(mirtlls)
mean(binbasmatrix[1:2, 1:2])-mean(mirtmeanprobs)
mean(binbasllmatrix[1:2, 1:2])-mean(mirtlls)