diff --git a/ccn2019.rev3.Rmd b/ccn2019.rev3.Rmd
index 468f63b..c0f15cd 100644
--- a/ccn2019.rev3.Rmd
+++ b/ccn2019.rev3.Rmd
@@ -1,3 +1,6 @@
+
+$P=\langle V,D,C,W \rangle$
+
 ```{r setup, message=FALSE, include=FALSE, paged.print=FALSE}
 #! ===============================================
 #! load required packages
@@ -94,7 +97,7 @@
 #! prepare data for modeling (remove na, etc)
 seqs <- seqs %>% 
   filter(!is.na(correct), !is.na(rt)) %>%
-  mutate(correct=factor(as.numeric(correct),labels=c("I","C")))
+  mutate(correct=factor(as.numeric(correct),labels=c("INCORRECT","CORRECT")))
   #mutate(correct=as.numeric(correct))
 
 #FIXME remove outcomes before dummy out the data and imputing missing values

diff --git a/ccn2019.rev3.Rmd b/ccn2019.rev3.Rmd
index 468f63b..c0f15cd 100644
--- a/ccn2019.rev3.Rmd
+++ b/ccn2019.rev3.Rmd
@@ -1,3 +1,6 @@
+
+$P=\langle V,D,C,W \rangle$
+
 ```{r setup, message=FALSE, include=FALSE, paged.print=FALSE}
 #! ===============================================
 #! load required packages
@@ -94,7 +97,7 @@
 #! prepare data for modeling (remove na, etc)
 seqs <- seqs %>% 
   filter(!is.na(correct), !is.na(rt)) %>%
-  mutate(correct=factor(as.numeric(correct),labels=c("I","C")))
+  mutate(correct=factor(as.numeric(correct),labels=c("INCORRECT","CORRECT")))
   #mutate(correct=as.numeric(correct))
 
 #FIXME remove outcomes before dummy out the data and imputing missing values
diff --git a/cnn2019-accuracy.R b/cnn2019-accuracy.R
new file mode 100644
index 0000000..42109e3
--- /dev/null
+++ b/cnn2019-accuracy.R
@@ -0,0 +1,93 @@
+#==================================================#
+# model the "accuract" column (a for global, and al for local accuracy)
+
+library(here)
+library(tidyverse)
+library(caret)
+library(inspectdf)
+
+
+load(here("data/nback_seqs.Rd"))
+
+set.seed(42)
+
+seqs.imputed <- seqs %>% 
+  filter(!is.na(correct), !is.na(rt)) %>%
+  select(-correct)
+
+inspect_num(seqs.imputed)
+
+seqs.dummy <- predict(dummyVars(~.,data=seqs.imputed),seqs.imputed)
+
+
+train_indexes <- createDataPartition(seqs.imputed$a,
+                                     times = 1,
+                                     p = 0.7,
+                                     list = F)
+
+train_data <- seqs.imputed[train_indexes,]
+test_data <- seqs.imputed[-train_indexes,]
+
+control <- trainControl(
+  method = "cv",
+  number = 5,
+  verboseIter = T
+)
+
+pls.new_model <- train(
+  a ~ t + l + s + v + n + tl + ll + sl + ul + vl,
+  data = train_data,
+  method = "pls",
+  preProcess = c("center","scale"),
+  trControl = control
+)
+
+pls.old_model <- train(
+  a ~ t + n + v,
+  data = train_data,
+  method = "pls",
+  preProcess = c("center","scale"),
+  trControl = control
+)
+
+
+pls.old_model
+pls.new_model
+varImp(pls.old_model)
+varImp(pls.new_model)
+
+trellis.par.set(caretTheme())
+densityplot(pls.new_model, pch = "|")
+densityplot(pls.old_model, pch = "|")
+
+resamps <- resamples(list(old = pls.old_model, new = pls.new_model))
+summary(resamps)
+dotplot(resamps, metric = "Rsquared")
+difValues <- diff(resamps)
+bwplot(difValues, layout=c(1,3))
+
+
+pls.new_train_predicted <- predict(pls.new_model, train_data, type="raw")
+pls.old_train_predicted <- predict(pls.old_model, train_data, type="raw")
+pls.new_predicted <- predict(pls.new_model, test_data, type="raw")
+pls.old_predicted <- predict(pls.old_model, test_data, type="raw")
+
+
+summary(pls.new_model)
+
+
+# SSE and RMSE
+
+SSE <- sum((test_data$a - pls.new_predicted)^2)    # sum of squared errors
+SST <- sum((test_data$a - mean(train_data$a))^2) # total sum of squares, remember to use training data here
+R_square <- 1 - SSE/SST
+SSE <- sum((test_data$a - pls.new_predicted)^2)
+RMSE <- sqrt(SSE/length(pls.new_predicted))
+
+
+SSE <- sum((test_data$a - pls.old_predicted)^2)
+R_square <- 1 - SSE/SST
+SSE <- sum((test_data$a - pls.old_predicted)^2)
+RMSE <- sqrt(SSE/length(pls.old_predicted))
+
+

diff --git a/ccn2019.rev3.Rmd b/ccn2019.rev3.Rmd
index 468f63b..c0f15cd 100644
--- a/ccn2019.rev3.Rmd
+++ b/ccn2019.rev3.Rmd
@@ -1,3 +1,6 @@
+
+$P=\langle V,D,C,W \rangle$
+
 ```{r setup, message=FALSE, include=FALSE, paged.print=FALSE}
 #! ===============================================
 #! load required packages
@@ -94,7 +97,7 @@
 #! prepare data for modeling (remove na, etc)
 seqs <- seqs %>% 
   filter(!is.na(correct), !is.na(rt)) %>%
-  mutate(correct=factor(as.numeric(correct),labels=c("I","C")))
+  mutate(correct=factor(as.numeric(correct),labels=c("INCORRECT","CORRECT")))
   #mutate(correct=as.numeric(correct))
 
 #FIXME remove outcomes before dummy out the data and imputing missing values
diff --git a/cnn2019-accuracy.R b/cnn2019-accuracy.R
new file mode 100644
index 0000000..42109e3
--- /dev/null
+++ b/cnn2019-accuracy.R
@@ -0,0 +1,93 @@
+#==================================================#
+# model the "accuract" column (a for global, and al for local accuracy)
+
+library(here)
+library(tidyverse)
+library(caret)
+library(inspectdf)
+
+
+load(here("data/nback_seqs.Rd"))
+
+set.seed(42)
+
+seqs.imputed <- seqs %>% 
+  filter(!is.na(correct), !is.na(rt)) %>%
+  select(-correct)
+
+inspect_num(seqs.imputed)
+
+seqs.dummy <- predict(dummyVars(~.,data=seqs.imputed),seqs.imputed)
+
+
+train_indexes <- createDataPartition(seqs.imputed$a,
+                                     times = 1,
+                                     p = 0.7,
+                                     list = F)
+
+train_data <- seqs.imputed[train_indexes,]
+test_data <- seqs.imputed[-train_indexes,]
+
+control <- trainControl(
+  method = "cv",
+  number = 5,
+  verboseIter = T
+)
+
+pls.new_model <- train(
+  a ~ t + l + s + v + n + tl + ll + sl + ul + vl,
+  data = train_data,
+  method = "pls",
+  preProcess = c("center","scale"),
+  trControl = control
+)
+
+pls.old_model <- train(
+  a ~ t + n + v,
+  data = train_data,
+  method = "pls",
+  preProcess = c("center","scale"),
+  trControl = control
+)
+
+
+pls.old_model
+pls.new_model
+varImp(pls.old_model)
+varImp(pls.new_model)
+
+trellis.par.set(caretTheme())
+densityplot(pls.new_model, pch = "|")
+densityplot(pls.old_model, pch = "|")
+
+resamps <- resamples(list(old = pls.old_model, new = pls.new_model))
+summary(resamps)
+dotplot(resamps, metric = "Rsquared")
+difValues <- diff(resamps)
+bwplot(difValues, layout=c(1,3))
+
+
+pls.new_train_predicted <- predict(pls.new_model, train_data, type="raw")
+pls.old_train_predicted <- predict(pls.old_model, train_data, type="raw")
+pls.new_predicted <- predict(pls.new_model, test_data, type="raw")
+pls.old_predicted <- predict(pls.old_model, test_data, type="raw")
+
+
+summary(pls.new_model)
+
+
+# SSE and RMSE
+
+SSE <- sum((test_data$a - pls.new_predicted)^2)    # sum of squared errors
+SST <- sum((test_data$a - mean(train_data$a))^2) # total sum of squares, remember to use training data here
+R_square <- 1 - SSE/SST
+SSE <- sum((test_data$a - pls.new_predicted)^2)
+RMSE <- sqrt(SSE/length(pls.new_predicted))
+
+
+SSE <- sum((test_data$a - pls.old_predicted)^2)
+R_square <- 1 - SSE/SST
+SSE <- sum((test_data$a - pls.old_predicted)^2)
+RMSE <- sqrt(SSE/length(pls.old_predicted))
+
+
diff --git a/cnn2019-correct.R b/cnn2019-correct.R
index a4e0b4c..b888bc1 100644
--- a/cnn2019-correct.R
+++ b/cnn2019-correct.R
@@ -41,6 +41,7 @@
   correct ~ t + l + s + v + n + tl + ll + sl + ul + vl,
   data = train_data,
   method = "pls",
+  preProcess = c("center","scale"),
   trControl = control
 )
 
@@ -48,6 +49,7 @@
   correct ~ t + n + v,
   data = train_data,
   method = "pls",
+  preProcess = c("center","scale"),
   trControl = control
 )
 
@@ -67,7 +69,7 @@
 bwplot(difValues, layout=c(1,3))
 
 
-pls.train_predicted1 <- predict(pls.model1, train_data, type="prob")
+pls.train_predicted1 <- predict(pls.model1, train_data, type="raw")
 pls.train_predicted2 <- predict(pls.model2, train_data, type="raw")
 pls.predicted1 <- predict(pls.model1, test_data, type="raw")
 pls.predicted2 <- predict(pls.model2, test_data, type="raw")