Exp1_dataprep.Rmd

---
title: Preparing data for modpd1
editor_options:
  chunk_output_type: console
output:
  pdf_document: default
  html_document: default
---

```{r prepare_variables,  echo = TRUE, warning=FALSE}
knitr::opts_chunk$set(echo = FALSE, warning=FALSE, message=FALSE, cache=TRUE)
library(readr)
library(data.table)

# Read and clean data set

## Data structure

# The EPrime experiment consists of two *condition* branches, *inclusion* and *exclusionAttitude*.
# Variable *Subject* indexes participant number.
# 
# Within each branch, there are the following phases/procedures:
# 
# - *preratings*: all face CSs (variable *ICS*) are rated on a slider (variable *SlideSlider.RESP*)
# - ListPair (list construction assigns USs to selected critical CSs): *SC*, *typUS*, *US[1..16]*
# - *conditioning* (presentation phase): variables *ICS*, *typUS*, *US[Trial]*
# - *TrainingInclusion* (1 trial for each of 8 cases): *correct*, *Slide2.RESP*, *Slide2.RT*
# - *PDB* (main PD procedure): variables *SC*/*CS* (paired CS: SC; filler CS: CSneg, CSpos), *SlidePDB.RESP*, *SlidePDB.RT* 
# - *postratings*: see *preratings*
# - *ValenceMem*: variables *ICS*, *Slide3.RESP*, *Slide3.RT* (1=pleasant, 2=unpleasant, 3=*don't know*)
# - *IdentityMem*: variables *ICS*, *SlideIm1.RESP*, *SlideIm1.RT*, *idMemSel* (selected US needs to be checked against ListPair data)


# The CSV file contains the data from all 129 participants collected in Louvain by Adrian Mierop (using EPrime V2).
# It contains a lot of useless columns that are removed in a first step.

# read data for Exp.1 (taken from Adrien Mierop via Dropbox download)
d <- read_delim("./data/modifiedPD.csv", ";", escape_double = FALSE, trim_ws = TRUE, guess_max = 20000)
## no problems

# remove useless columns
d$Group <- d$Block <- d$BlockList <- d$BlockList.Cycle <- d$BlockList.Sample <- d$`Running[Block]` <- d$conditioning.Cycle <-  NULL
d$`Procedure[SubTrial]` <- d$`Running[SubTrial]` <- d$IdentityMem <- d$IdentityMem.Cycle <- d$ListPair.Cycle <- d$PDB.Cycle <- NULL
d$postratings.Cycle <- d$preratings.Cycle <- d$ValenceMem.Cycle <-  NULL

# easy alias for important columns
d$task <- d$`Running[Trial]`
d$rating <- d$SlideSlider.RESP
d$vm.resp <- d$Slide3.RESP
d$vm.rt <- d$Slide3.RT
d$im.resp <- d$SlideIm1.RESP
d$im.rt <- d$SlideIm1.RT
d$pd.resp <- d$SlidePDB.RESP
d$pd.rt <- d$SlidePDB.RT


# Technical Exclusions

# exclude practice-run Subject 1
d <- subset(d, Subject>1)
# had problems with CS/US assignment
d <- subset(d, Subject != 67)
d <- subset(d, Subject != 71)
d <- subset(d, Subject != 60)
d <- subset(d, Subject != 78)
# was run in both inclusion and exclusion conditions
d <- subset(d, Subject != 87) 


# Prepare variables

## for a given CS, read out the paired USs

list <- subset(d, task=="ListPair", select=c(Condition, Subject, SC, typUS, `US[SubTrial]`))
names(list) <- c("Condition","Subject","ICS","typUS","US")
lu <- unique(subset(list, select=-US))

## evaluative ratings

pre <- subset(d, task=="preratings", select = c(Condition, Subject, ICS, rating))
names(pre) <- c("Condition","Subject","ICS","prerating")
post <- subset(d, task=="postratings", select = c(Condition, Subject, ICS, rating))
names(post) <- c("Condition","Subject","ICS","postrating")

## memory for US valence and US identity

vm <- subset(d, task=="ValenceMem", select = c(Condition, Subject, ICS, vm.resp, vm.rt))
im <- subset(d, task=="IdentityMem", select = c(Condition, Subject, ICS, im.resp, im.rt, US1, US2, US3, US4, US5, US6, US7, US8, US9, US10, US11, US12, US13, US14, US15, US16, USi1, USi2, USi3, USi4, USi5,USi6, USi7, USi8, USi9, USi10, USi11, USi12, USi13, USi14, USi15, USi16))


### compute IDmem correctness

l <- merge(im, list, by=c("Condition","Subject", "ICS"), all=FALSE)
l[6:21] <- lapply(l[6:21], function(x) substr(x,7, 11))
l[,"US"] <- NULL
l <- unique(l)

l$selec <- paste('USi', l$im.resp, sep='')

l <- data.table(l)

l[selec %in% "USi1", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi2", c("USi1", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi3", c("USi2", "USi1", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi4", c("USi2", "USi3", "USi1", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi5", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi6", c("USi2", "USi3", "USi4", "USi5", "USi1", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi7", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi1", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi8", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi1", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi9", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi1", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi10", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi1", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi11", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi1", "USi12", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi12", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi1", "USi13", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi13", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi1", 'USi14', "USi15", "USi16")  := NA ]
l[selec %in% "USi14", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi1', "USi15", "USi16")  := NA ]
l[selec %in% "USi15", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi1", "USi16")  := NA ]
l[selec %in% "USi16", c("USi2", "USi3", "USi4", "USi5", "USi6", "USi7", "USi8", "USi9", "USi10", "USi11", "USi12", "USi13", 'USi14', "USi15", "USi1")  := NA ]
l[, selUS := sum(USi1, USi2, USi3, USi4, USi5, USi6, USi7, USi8, USi9, USi10, USi11, USi12, USi13, USi14, USi15, USi16, na.rm= TRUE), 1:nrow(l)]
l[selUS == US1 | selUS == US2| selUS == US3| selUS == US4| selUS == US5| selUS == US6| selUS == US7| selUS == US8, im.correct := 1]
l[selUS == US9 | selUS == US10| selUS == US11| selUS == US12| selUS == US13| selUS == US14| selUS == US15| selUS == US16, im.correct := 0]
l$im.correct.b <- ifelse(l$im.correct==1, TRUE, FALSE)
lc <- aggregate(x=l$im.correct.b, by=list(Condition=l$Condition, Subject=l$Subject, ICS=l$ICS), FUN=any)
names(lc) <- c("Condition","Subject","ICS","im.correct")
im <- merge(subset(im, select=c(Condition, Subject, ICS, im.resp, im.rt)), lc, by=c("Condition","Subject","ICS"))


## pd training accuracy

pdtrain <- subset(d, !is.na(correct), select = c(Condition, Subject, TrainingInclusion.Sample, correct, Slide2.RESP, Slide2.RT))
pdtrain$ACC <- ifelse(pdtrain$correct == pdtrain$Slide2.RESP, 1, 0)
#write.csv(pdtrain, file="./data/modPD1_pdtrain.csv")
pdc <- aggregate(pdtrain$ACC, by=list(Condition=pdtrain$Condition, Subject=pdtrain$Subject), FUN=mean)
names(pdc) <- c("Condition","Subject","pdtrain.correct")


## pd responses

pd <- subset(d, task=="PDB", select=c(Condition, Subject, ICS, CS, SC, CSpos, CSneg, pd.resp, pd.rt))
pd$typCS <- ""
pd$typCS[pd$ICS == pd$SC] <- "neutral"
pd$typCS[pd$ICS == pd$CS & !is.na(pd$CSpos)] <- "pos"
pd$typCS[pd$ICS == pd$CS & !is.na(pd$CSneg)] <- "neg"

pd$pdset <- NA
pd$pdset[pd$pd.resp %in% c(1,2)] <- 1
pd$pdset[pd$pd.resp %in% c(3,4)] <- 2
pd$pdset <- factor(pd$pdset, labels = c("Memory", "Attitude"))

pd$pdval <- NA
pd$pdval[pd$pd.resp %in% c(1,3)] <- 1
pd$pdval[pd$pd.resp %in% c(2,4)] <- 2
pd$pdval <- factor(pd$pdval, labels=c("pleasant", "unpleasant"))


## merge variables (and compute some)

dd <- merge(lu, subset(pd, select=c(Condition, Subject, ICS, typCS, pdset, pdval, pd.rt)), by=c("Condition","Subject","ICS"))

## ratings
dd <- merge(dd, pre, by=c("Condition","Subject","ICS"))
dd <- merge(dd, post, by=c("Condition","Subject","ICS"))

### valmem
dd <- merge(dd, vm, by=c("Condition","Subject","ICS"))
dd$vm.correct <- NA
dd$vm.correct[dd$vm.resp==1 & dd$typUS=="pos"] <- TRUE
dd$vm.correct[dd$vm.resp==2 & dd$typUS=="neg"] <- TRUE
dd$vm.correct[dd$vm.resp==1 & dd$typUS=="neg"] <- FALSE
dd$vm.correct[dd$vm.resp==2 & dd$typUS=="pos"] <- FALSE

### idmem
dd <- merge(dd, subset(im, select=c(Condition, Subject, ICS,im.correct, im.rt)), by=c("Condition","Subject","ICS"))

dd <- merge(dd, pdc)

## create factors

dd$Condition <- relevel(factor(dd$Condition, labels=c("Exclusion", "Inclusion")), ref="Inclusion")
dd$Subject <- as.factor(dd$Subject)
dd$ICS <- as.factor(dd$ICS)
dd$typCS <- as.factor(dd$typCS)
dd$typUS <- as.factor(dd$typUS)
dd$vm.resp <- factor(dd$vm.resp, labels=c("pleasant", "unpleasant", "dont know")) # (1=pleasant, 2=unpleasant, 3=*don't know*)
dd$vm.correct <- as.factor(dd$vm.correct)
dd$im.resp <- NULL
dd$im.correct <- as.factor(dd$im.correct)

d1 <- dd

### exclude Ss with deficient instruction understanding
#d1 <- subset(d1, Condition !="Exclusion")
#n1 <- length(unique(d1$Subject))

length(unique(d1$Subject))
d1 <- subset(d1, pdtrain.correct>=.75)
length(unique(d1$Subject))

## write to disk

save(d1, file="./data/modPD1.RData")
write.csv(d1, file="./data/Exp1.csv")

```