# Combining data from five EXPeriments
# Kliegl et al. (2009, PsycholRes)
# May 4, 2007:
# Key to variables
# ...cuePos -1 links, 1 rechts
# ...cueTyp: 
# ...tarPos: Target pos -1 links, -1 rechts
# ...tFixEin: FCI
# ...tTarEin: CTI
# ...numBinSac: n of binocular sacc (incl. micro)
# ...sacType: 1 micro, 2 sacc response (not here) ,  3 manual response
# ...sacOnset: CSacI (- before cue, + after cue)
# ...sacOffset:  like sacOnset
# ...sacDur:  saccade duration
# ...sacDelay: interocular differece of sacc onset
# ...sacVPeak: peak velocity
# ...sacDist: effective distance
# ...sacAngle1: direction (-pi to + pi) for sacDist
# ...sacAmp: maximum distance
# ...sacAngle2: direction (-pi to + pi) for sacAmp
# ...sacxOset, sacyOnset, sacxOffset, sacyOffset: drift from fix point
# ...reaRT: manual RT
# ...reaCor: 0 =wrong 1=falsch

# May 14, 2007
# revision: June 13, 2007
# revision: January 21, 2008
#

# NOTE THE FOLLOWING VARIABLES ARE AVAILABLE IN THE DATA FRAME MM.rda
######### EXPerimental codes ##################
# ... cv:   invalid = 0, valid = 1
# d$cv <- ifelse(d$cuePos==d$tarPos, 1, 0)
# d0$cv <- ifelse(d0$cuePos==d0$tarPos, 1, 0)

# ... cue direction: left = 0, right = 1
# d$cd <- ifelse(d$cuePos==-1, 0, 1)
# d0$cd <- ifelse(d0$cuePos==-1, 0, 1)

# ... target position: left = 0, right = 1
# d$tpos <- ifelse(d$tarPos==-1, 0, 1)
# d0$tpos <- ifelse(d0$tarPos==-1, 0, 1)

# ... effect coding for cv, mtc
# d$cv.e   <- d$cv   - 0.5
# d$CV <- as.factor(d$cv)
# levels(d$CV) <- c("invalid", "valid")
# d0$cv.e   <- d0$cv   - 0.5
# d0$CV <- as.factor(d0$cv)
# levels(d0$CV) <- c("invalid", "valid")

########## MS characteristics
# ms time of occurrence
# d$mt <- d$sacOnset
# d0$mt <- NA
# d$mt_r <- d$mt - d$tTarEin 
# d0$mt_r <- NA

# ... ms direction: left = 0, right = 1
# d$md <- ifelse(d$sacAngle1 <= -pi/2 | d$sacAngle1 > pi/2, 0, 1)
# d0$md <- NA

# ... ms congruency with cue:    incongruent = 0, congruent = 1
# d$mcc <- ifelse(d$md==d$cd, 1, 0)
# d0$mcc <- NA

# ... ms congruency with target: incongruent = 0, congruent = 1
# d$mtc <- ifelse(d$md==d$tpos, 1, 0)
# d0$mtc <- NA


# d$mtc.e  <- d$mtc  - 0.5
# d$MTC <- as.factor(d$mtc)
# levels(d$MTC) <- c("incongruent", "congruent")
# d0$mtc.e <- NA
# d0$MTC <- NA


# d$mcc.e  <- d$mcc  - 0.5
# d$MCC <- as.factor(d$mcc)
# levels(d$MCC) <- c("incongruent", "congruent")
# d0$mcc.e <- NA
# d0$MCC <- NA

################################################
# dependent variables and transformations
# d$rt <- d$reaRT
# d0$rt <- d0$reaRT

# d$lrt <- log(d$rt)
# d0$lrt <- log(d0$rt)

##################################################
# generate compatible data frames
# d0 <- d0[, c("ID", "EXP", "TYP", "CV", "MTC", "rt", "lrt", "mt", "mt_r", "sacAmp", "sacVPeak", "mn", "cv.e", "mtc.e", "tpos")]
# d <-   d[, c("ID", "EXP", "TYP", "CV", "MTC", "rt", "lrt", "mt", "mt_r", "sacAmp", "sacVPeak", "mn", "cv.e", "mtc.e", "tpos")]
# d <- rbind(d, d0)
# d[] <- lapply(d,function(x) x[drop=TRUE])

# Save MMVV
# save(d, file="data/MM.rda")

# EXTRAS
library(ggplot2)
library(lme4)

# cue validity effect
rm(list=ls())
load("MM.rda")

d.rs <- melt(d, id=c("ID", "EXP", "TYP", "CV", "MTC"), measure=c("rt", "lrt", "mn"), subset=TYP !="first MS")
(table.id <- cast(d.rs, ID + EXP + CV ~ variable, function(x) c(M=mean(x), N=length(x))))
summary(aov(rt_M ~ EXP*CV, data=table.id))
summary(aov(mn_M ~ EXP*CV, data=table.id))
summary(aov(mn_N ~ EXP*CV, data=table.id))

# PART 1: EFFECT OF N OF MS IN TRIAL ON RT
# include "No MS" trials, i.e. leave out MTC, use only first of several MS
ix <- which(d$TYP !="last MS")
d.sub <- d[ix, ]
d.sub[] <- lapply(d.sub,function(x) x[drop=TRUE])

d.sub$TYP <- relevel(d.sub$TYP, ref=3)
d.sub$TYP2 <- C(d.sub$TYP, matrix(c(-2, +1, +1,  0, -1, 1), 3, 2), 2)
contrasts(d.sub$TYP2)
d.sub$MTC <- relevel(d.sub$MTC, ref="congruent")       # ref is "target-congruent MS"

print(m0 <- lmer(rt ~ (EXP+CV+TYP2)^2 + (1|ID), data=d.sub), cor=FALSE)
print(m0.VV <- lmer(rt ~  CV*TYP2 + (1|ID), data=d.sub, subset=EXP=="VV"), cor=FALSE)
print(m0.VA <- lmer(rt ~  CV*TYP2 + (1|ID), data=d.sub, subset=EXP=="VA"), cor=FALSE)
# Main message: MS for neutral cues slightly increase RT, for invalid cues they slightly reduce cost, and for valid cues they increase benefit!

d.sub.rs <- melt(d.sub, id=c("ID", "EXP", "TYP", "CV", "MTC"), measure=c("rt", "lrt", "mn"))
(table <- cast(d.sub.rs, EXP + CV + TYP ~ variable, function(x) c(M=mean(x), N=length(x))))
levels(table$TYP) <- c("0 MS", "1 MS", "2+ MS")
qplot(x=CV, y=rt_M, data=table, colour=TYP, group=TYP, facets= EXP ~ TYP, geom=c("point", "line"))
qplot(x=CV, y=lrt_M, data=table, colour=TYP, group=TYP, facets=EXP ~ TYP, geom=c("point", "line"))

# PART 1: EFFECT OF N OF MS IN TRIAL ON RT
ix <- which(d$TYP == "no MS")
d.sub12 <- d[-ix, ]
d.sub12[] <- lapply(d.sub12,function(x) x[drop=TRUE])

# include MTC as factor, leave out "No MS" trials; combine for nesting within TYP?, separate LMEs for MTC of first and last MS
print(m1 <- lmer(rt ~  (EXP+CV+TYP+MTC)^3 + (1|ID), data=d.sub12, subset=TYP !="last MS" ), cor=FALSE)
print(m2 <- lmer(rt ~  (EXP+CV+TYP+MTC)^3 + (1|ID), data=d.sub12, subset=TYP !="first MS"), cor=FALSE)

d.sub12.rs <- melt(d.sub12, id=c("ID", "EXP", "TYP", "CV", "MTC"), measure=c("rt", "lrt", "mn"))
(table12 <- cast(d.sub12.rs, EXP + CV + TYP + MTC ~ variable, function(x) c(M=mean(x), N=length(x))))
table12$CV <- relevel(table12$CV, ref="invalid")
levels(table12$TYP) <- c("single MS", "first MS", "last MS")
qplot(x=CV, y=rt_M, colour=MTC, group=MTC,  facets= EXP ~ TYP, data=table12, geom = c("point", "line"))
qplot(x=CV, y=lrt_M, colour=MTC, group=MTC, facets= EXP ~ TYP, data=table12, geom = c("point", "line"))
#

# Extra for plots
#	   function(x) c(M=mean(x), SE=1.96*sd(x)/sqrt(length(x)), N=length(x) ))
#p <- p +geom_errorbar(aes(max=M+SE, min=M-SE), width=0.1)
#p