##########      Am I sure that happened, or am I just making it up? Investigating false memories after
##########      alcohol-induced memory blackouts in sober young adults.     Jackson & Dering, 2022

##########                          Analysis Code      ##################


###### Contents

#1. dPrime Between Groups, including split by encoding
#2. C Between Groups, including split by encoding
#3. dPrime Within Group, including split by encoding
#4. C Within Group, including split by encoding
#5. RKG by Hits (%), Between Groups
#6. False Alarms by Rel/Unrel, Between Groups
#7. RKG by Hits (%) Within MBO Group
#8. FA by Rel/Unrel, within MBO Group
#9. Sleep and dPrime, plus Bayesian
#10.Sleep and FA, plus Bayesian 

############################################################
# Load packages
library(Hmisc)
library(nlme)
library(reshape)
library(multcomp)               
library(pastecs)
library(psych)
library(Matrix)
library(lme4)
library(gridExtra)
library(car)
library(readxl)


#####################################################################################################################
##########################    Whole Group - dPrime    ##############################################################
# Load data.  All data numbers in % 

DRM_DPrime <- read_excel("~ ..........DRM_DPrime_Between.xlsx")
DRM_DPrime$Participant <- as.numeric(as.integer(DRM_DPrime$Participant))
str(DRM_DPrime)

# Separate data sets for dPrime and C.
DRM_dp1 <- DRM_DPrime[1:106, c("Participant","Group", "Alcohol", "dFree","dSerial")]
DRM_dp <- as.data.frame(DRM_dp1)
str(DRM_dp)

DRM_c1 <- DRM_DPrime[1:106, c("Participant", "Group", "Alcohol", "cFree", "cSerial")]
DRM_c <- as.data.frame(DRM_c1)
str(DRM_c)

###################################################################################################
###########################    Begin analysis of dPrime, between groups    ######################################

# Reshape data from wide into long format

DRM_DP <- melt.data.frame(DRM_dp, id.vars = c("Participant", "Group", "Alcohol"), 
                          measure.vars = c("dFree", "dSerial")) 
View(DRM_DP)

names(DRM_DP)<- c("Participant", "Group", "Alcohol", "Condition", "dPrime")
str(DRM_DP)

# Convert the 'Condition" variable into a column with 2 levels (Free, Serial). Change Alcohol and Group into factors.

DRM_DP$Condition <- gl(2,106, labels = c("Free", "Serial"))
DRM_DP$Group<-as.factor(DRM_DP$Group)
DRM_DP$Alcohol <- as.factor((DRM_DP$Alcohol))
str(DRM_DP)
View(DRM_DP)

############   Descriptives     ##########

# Descriptives by group x alcohol
by(DRM_DP$dPrime, list(DRM_DP$Alcohol), stat.desc, basic=F)

############   Begin modelling analysis     ##########

# Assess need for multilevel model

interceptdrmdp <- gls(dPrime ~ 1, data = DRM_DP, method = "ML", )
randominterceptdrmdp <- lme(dPrime ~1, data= DRM_DP, random = ~1|Group, method = "ML")
anova(interceptdrmdp, randominterceptdrmdp)

# Not signficant therefore no need to do multilevel model.

BeforevsAfter<-c(1,-1)
ControlvsMBO<-c(-1,1)
FreevsSerial <-c(-1,1)

contrasts(DRM_DP$Alcohol)<-cbind(BeforevsAfter)
contrasts(DRM_DP$Group) <- cbind(ControlvsMBO)
contrasts(DRM_DP$Condition) <-cbind(FreevsSerial)

# Check contrasts

DRM_DP$Condition
DRM_DP$Group
DRM_DP$Alcohol

# Create baseline mixed effects model.  This is a Factorial Repeated-Measures Design 

basedrmdp <- lme(dPrime ~ 1, random = ~1|Participant/Group/Alcohol/Condition, 
                 data = DRM_DP, method = "ML")   # This model predicts outcome by intercept, and the random part says that
                                                 # scores for each participant are within condition and group variables.
summary(basedrmdp)

# Add main effects by updating base model.
drmdp1 <- update(basedrmdp, .~. + Group)  # Add group as a predictor of dPrime
drmdp2 <- update(drmdp1, .~. + Alcohol)   # Add Alcohol as a predictor
drmdp3 <- update(drmdp2, .~. + Condition) #Add encoding as a predictor

# Add interaction(s)
drmdp4 <- update(drmdp3, .~. +Group:Alcohol)
drmdp5 <- update(drmdp4, .~. +Group:Condition)
drmdp6 <- update(drmdp5, .~. +Alcohol:Condition)
drmdp7 <- update(drmdp6, .~. +Group:Alcohol:Condition)
# Compare models
anova(basedrmdp, drmdp1, drmdp2, drmdp3, drmdp4, drmdp5, drmdp6, drmdp7)

#Model drmdp2 is significant, suggesting that only alcohol is a predictor of dPrime

summary(drmdp2)

# Write function to calculate effect size
rcontrast <- function(t,df)
{r<- sqrt(t^2/(t^2 + df))
print(paste("r = ", r))
}
rcontrast(-4.256832, 52)  # effect size for main effect of alcohol.  Use t-value and df from model summary.
# r =  0.50835095181327

###################################################################################################################
######################   c Between Groups  #######################################

DRM_c1 <- DRM_DPrime[1:106, c("Participant", "Group", "Alcohol", "cFree", "cSerial")]
DRM_c <- as.data.frame(DRM_c1)
str(DRM_c)


# Reshape data from wide into long format

DRM_C <- melt.data.frame(DRM_c, id.vars = c("Participant", "Group", "Alcohol"), 
                         measure.vars = c("cFree", "cSerial")) 
View(DRM_C)

names(DRM_C)<- c("Participant", "Group", "Alcohol", "Condition","c")

# Turn Group and Alcohol into factors, turn Condition into column with 2 levels (Free, Serial).

DRM_C$Condition <- gl(2,106, labels = c("Free","Serial"))
DRM_C$Alcohol <-as.factor(DRM_C$Alcohol)
DRM_C$Group<-as.factor(DRM_C$Group)
str(DRM_C)


# Get descriptive statistics 
#I.Descriptives - Condition, all participants 
by(DRM_C$c, list(DRM_C$Condition), stat.desc, basic=F)
# Descriptives by group x alcohol x encoding condition
by(DRM_C$c, list(DRM_C$Group, DRM_C$Alcohol,DRM_C$Condition), stat.desc, basic=F)
# Descriptives by alcohol x encoding condition
by(DRM_C$c, list(DRM_C$Alcohol,DRM_C$Condition), stat.desc, basic=F)

############   Begin modelling analysis     ##########

# Assess need for multilevel model

interceptdrmc <- gls(c ~ 1, data = DRM_C, method = "ML")
randominterceptdrmc <- lme(c ~1, data= DRM_C, random = ~1|Group, method = "ML")
anova(interceptdrmc, randominterceptdrmc)

# Not signficant therefore no need to do multilevel model.

# Set contrasts before building models. 

BeforevsAfter<-c(-1,1)
ControlvsMBO<-c(-1,1)
FreevsSerial<-c(-1,1)

contrasts(DRM_C$Condition)<-cbind(FreevsSerial)
contrasts(DRM_C$Group) <- cbind(ControlvsMBO)
contrasts(DRM_C$Alcohol)<-cbind(BeforevsAfter)

# Check contrasts are correct

DRM_C$Condition
DRM_C$Group
DRM_C$Alcohol

# Create baseline mixed effects model.  This is a Factorial Repeated-Measures Design 

basedrmc <- lme(c ~ 1, random = ~1|Participant/Group/Alcohol/Condition, 
                data = DRM_C, method = "ML")   # This model predicts outcome by intercept, and the random part says that
                                               # scores for each participant are within condition and group variables.
summary(basedrmc)

# Add main effects by updating base model.
drmc1 <- update(basedrmc, .~. + Group)  # Add group as a predictor
drmc2 <- update(drmc1, .~. +Alcohol)    # Add alcohol as a predictor
drmc3 <- update(drmc2, .~. +Condition)  # Add encoding as a factor
# Add interaction(s)
drmc4 <- update(drmc3, .~. +Group:Alcohol)
drmc5 <- update(drmc4, .~. +Group:Condition)
drmc6 <- update(drmc5, .~. +Alcohol:Condition)
drmc7 <- update(drmc6, .~. +Group:Alcohol:Condition)

# Compare models
anova(basedrmc, drmc1, drmc2, drmc3, drmc4, drmc5, drmc6, drmc7)

#drmc6 significant 
summary(drmc6)

rcontrast(7.828691, 51) #r = 0.738791175811038 main effect of alcohol
rcontrast(-3.33658,103) #r =  0.312317536815095 interaction between alcohol and encoding condition

###################################################################################################################
#######################    dPrime Within MBO Group     ############################################################

# Load data.  All data numbers in % 
DRM_DPrimeM <- read_excel(".........\\DRM_DPrime_Within.xlsx")#, header = TRUE, stringsAsFactors = FALSE)


DRM_DPrimeM$Participant <- as.numeric(as.integer(DRM_DPrimeM$Participant))
str(DRM_DPrimeM)
View(DRM_DPrimeM)

# Want to create separate data sets for dPrime and C.
DRM_dpM1 <- DRM_DPrimeM[1:69, c("Participant", "Alcohol","dFree", "dSerial")]
DRM_dpM <- as.data.frame(DRM_dpM1)
str(DRM_dpM)

DRM_cM1 <- DRM_DPrimeM[1:69, c("Participant", "Alcohol", "cFree", "cSerial")]
DRM_cM <- as.data.frame(DRM_cM1)
str(DRM_cM)

# Reshape dprime data from wide into long format

DRM_DPM <- melt.data.frame(DRM_dpM, id.vars = c("Participant", "Alcohol"), 
                           measure.vars = c("dFree", "dSerial")) 
View(DRM_DPM)

names(DRM_DPM)<- c("Participant", "Alcohol", "Encoding", "dPrime")

# Convert Alcohol into a factor.

DRM_DPM$Encoding <- gl(2,69, labels = c("dFree", "dSerial"))
DRM_DPM$Alcohol <- as.factor(DRM_DPM$Alcohol)
str(DRM_DPM)

# Get descriptive statistics 
#I.Descriptives - Alcohol
by(DRM_DPM$dPrime, list(DRM_DPM$Alcohol, DRM_DPM$Encoding), stat.desc, basic=F)

############   Begin modelling analysis     ##########

# set contrasts before building models. 

AftervsAfterMBO<-c(-1,1,0) # Compare before alcohol to after alcohol, ignores after MBO
BeforevsAfter<-c(1,0,-1)   # Compare after alcohol to after MBO, ignores before
FreevsSerial <- c(-1,1)

contrasts(DRM_DPM$Alcohol)<-cbind(BeforevsAfter, AftervsAfterMBO)
contrasts(DRM_DPM$Encoding)<-cbind(FreevsSerial)
#Check contrasts are correct
DRM_DPM$Alcohol
DRM_DPM$Encoding


# Create baseline mixed effects model.  This is a Factorial Repeated-Measures Design 

basedpM <- lme(dPrime ~ 1, random = ~1|Participant/Alcohol/Encoding, 
               data = DRM_DPM, method = "ML", na.action = na.exclude)   # This model predicts outcome by intercept, and the random part says that
                                                                        # scores for each participant is within condition variables.
summary(basedpM)

# Add main effects by updating base model.
drmdpM <- update(basedpM, .~. +Alcohol)  #Add alcohol as a predictor
drmdpM1 <- update(drmdpM, .~. +Encoding) #Add encoding type as a predictor
drmdpM2 <- update(drmdpM1, .~. +Alcohol:Encoding) # Add interaction

# Compare models
anova(basedpM, drmdpM, drmdpM1, drmdpM2)

summary(drmdpM) 

#Bonferroni 
pairwise.t.test(DRM_DPM$dPrime, DRM_DPM$Alcohol,paired = TRUE, p.adjust.method = "bonferroni")

#         After AfterMBO
#AfterMBO 0.999 -       
#Before   0.006 0.231    

##################################################################################################################
#######################    c Within MBO Group     ############################################################

# Reshape data from wide into long format

DRM_CM <- melt.data.frame(DRM_cM, id.vars = c("Participant", "Alcohol"), 
                          measure.vars = c("cFree", "cSerial")) 

View(DRM_CM)
names(DRM_CM)<- c("Participant", "Alcohol", "Encoding", "c")

# Now we have to convert alcohol and encoding Condition into factors.

str(DRM_CM)
DRM_CM$Alcohol <- as.factor(DRM_CM$Alcohol)
DRM_CM$Encoding <- gl(2,69, labels = c("cFree", "cSerial"))
str(DRM_CM)

# Get descriptive statistics 
#I.Descriptives - Condition
by(DRM_CM$c, list(DRM_CM$Alcohol,DRM_CM$Encoding ), stat.desc, basic=F)

############   Begin modelling analysis     ##########

# Set contrasts before building models.  Need to set these for Alcohol

BeforevsAfter<-c(1,0,-1)   # Compare before alcohol to after alcohol, ignores after MBO
AftervsAfterMBO<-c(-1,1,0) # Compare after alcohol to after MBO, ignores before
FreevsSerial <-c(-1,1)

contrasts(DRM_CM$Alcohol)<-cbind(BeforevsAfter,AftervsAfterMBO)
contrasts(DRM_CM$Encoding)<-cbind(FreevsSerial)

#Check contrasts are correct
DRM_CM$Alcohol
DRM_CM$Encoding

# Create baseline mixed effects model.  This is a Factorial Repeated-Measures Design

basecM <- lme(c ~ 1, random = ~1|Participant/Alcohol/Encoding, 
              data = DRM_CM, method = "ML", na.action = na.exclude)   # This model predicts outcome by intercept, and the random part says that
                                                                      # scores for each participant is within condition variables.
summary(basecM)

# Add main effects by updating base model.
drmcM <- update(basecM, .~. +Alcohol)   # Add alcohol as a predictor
drmcM1 <-update (drmcM, .~. +Encoding)  # Add encoding as a predictor
drmcM2 <-update (drmcM1, .~. +Alcohol:Encoding) # Add interaction as a predictor

# Compare models
anova(basecM, drmcM, drmcM1, drmcM2)

summary(drmcM2)

#Bonferroni
pairwise.t.test(DRM_CM$c, DRM_CM$Alcohol, paired = TRUE, p.adjust.method = "bonferroni")

#          After   AfterMBO
#AfterMBO   0.0684  -       
#Before     2.3e-05 0.0075  

###########################################################################################################################################
######################################   RKG by Hits, Between Groups   ####################################################################

# Load data.  All data numbers in % 

DRM_RKG_Between <- read_excel("~........\\RKG_Between.xlsx")
DRM_RKG_Between$Participant <- as.numeric(as.integer(DRM_RKG_Between$Participant))
DRM_RKG_Between <- as.data.frame(DRM_RKG_Between)
str(DRM_RKG_Between)

# Create seperate data set for :-
# Hits, between groups.  Totals as perentages, RKG's as proportions (%) of totals
DRM_RKGHitsBetween <- DRM_RKG_Between[1:106, c("Participant", "Group", "Alcohol", "HitTotal", "HitGuess", "HitKnow", "HitRemember")]
View(DRM_RKGHitsBetween)
str(DRM_RKGHitsBetween)

# RKG Hits
DRM_RKGHit <- melt.data.frame(DRM_RKGHitsBetween, id.vars = c("Participant", "Group", "Alcohol"), 
                              measure.vars = c("HitGuess", "HitKnow", "HitRemember"))
str(DRM_RKGHit)

names(DRM_RKGHit)<- c("Participant", "Group", "Alcohol", "RKG", "Hits")

DRM_RKGHit$Alcohol<-as.factor(DRM_RKGHit$Alcohol)
DRM_RKGHit$Group <-as.factor(DRM_RKGHit$Group)
DRM_RKGHit$RKG <- as.factor(DRM_RKGHit$RKG)
DRM_RKGHit$Participant <-as.numeric(DRM_RKGHit$Participant)
str(DRM_RKGHit)


############################################################################################
##############   Total Hits by RKG  ####################### 

# Set contrasts
BeforevsAfter<-c(1,-1)
ControlvsMBO<-c(-1,1)
KnowvsGuess<-c(1,-1,0)
KnowvsRemember <-c(0,-1,1)

contrasts(DRM_RKGHit$Alcohol)<-cbind(BeforevsAfter)
contrasts(DRM_RKGHit$Group) <- cbind(ControlvsMBO)
contrasts(DRM_RKGHit$RKG)<- cbind(KnowvsGuess, KnowvsRemember)

# Check contrasts are correct
DRM_RKGHit$Alcohol
DRM_RKGHit$Group
DRM_RKGHit$RKG

# Create baseline mixed effects model.  This is a Factorial Repeated-Measures Design 

basedrmhits <- lme(Hits ~ 1, random = ~1|Participant/Group/Alcohol/RKG, 
                   data = DRM_RKGHit, method = "ML")   # This model predicts outcome by intercept, and the random part says that
                                                       # scores for each participant are within condition and group variables.
summary(basedrmhits)

# Add main effects by updating base model.
drmhrkg1 <- update(basedrmhits, .~. + Group)   # Add group as a predictor of MeanACC
drmhrkg2 <- update(drmhrkg1, .~. + Alcohol)    # Add alcohol as a predictor
drmhrkg3 <- update(drmhrkg2, .~. + RKG)        # Add rkg as a predictor

# Add interaction(s)
drmrkgh4 <- update(drmhrkg3, .~. +Group:Alcohol)
drmrkgh5 <- update(drmrkgh4, .~. +Group:RKG)
drmrkgh6 <- update(drmrkgh5, .~. +Alcohol:RKG)
drmrkgh7 <- update(drmrkgh6, .~. +Group:Alcohol:RKG)

anova(basedrmhits, drmhrkg1, drmhrkg2, drmhrkg3, drmrkgh4, drmrkgh5, drmrkgh6, drmrkgh7) #same as previous output

# drmrkgh6 is best
summary(drmrkgh6) #same as previous
rcontrast(-17.46565,206) #Know vs Guess, r =  0.772597878568171
rcontrast(17.94658,206) #Know vs Remember, r =  0.780965679809749
rcontrast(-3.062,206) #Group vs Know/Remember, r =  0.208644379542681
rcontrast(-3.63214,206) #Alcohol vs Know/Remember,r =  0.245329511124844
#Bonferroni
pairwise.t.test(DRM_RKGHit$Hits, DRM_RKGHit$RKG, paired = FALSE, p.adjust.method = "bonferroni")

#           HitGuess HitKnow
#HitKnow     <2e-16   -      
#HitRemember <2e-16   <2e-16 

# Get descriptive statistics 
#I.Descriptives - Condition, all participants 
by(DRM_RKGHit$Hits, list(DRM_RKGHit$RKG, DRM_RKGHit$Group), stat.desc, basic=F)
  
#2.Descriptives - RKG by Alcohol, all participants  
by(DRM_RKGHit$Hits, list(DRM_RKGHit$RKG, DRM_RKGHit$Alcohol), stat.desc, basic=F)

#3.Descriptives - RKG by Alcohol by Group, all participants 
by(DRM_RKGHit$Hits, list(DRM_RKGHit$RKG, DRM_RKGHit$Alcohol, DRM_RKGHit$Group), stat.desc, basic=F)


###############################################################################################################################
########   Between Groups, False Alarm's  #####################################  

# FA's, between groups.  Total FAs as %, RKG's as proportions of related and unrelated totals
DRM_RKGFABDRM_RKGFABetween <- DRM_RKG_Between[1:106, c("Participant", "Group", "Alcohol", "FATotal", "RelFAGuess", "RelFAKnow", "RelFARemember", "FARelTotal",
                                                       "UnrelFAGuess", "UnrelFAKnow", "UnrelFARemember", "UnrelFATotal")]

View(DRM_RKGFABetween)

## Prepare the datasets

# Related vs Unrelated FA's - this takes the total Rel and Unrel FA's from above dataset, leaves out the RKG split
DRM_RKGRUFA <- melt.data.frame(DRM_RKGFABetween, id.vars = c("Participant", "Group", "Alcohol"),
                               measure.vars = c("FARelTotal", "UnrelFATotal"))
str(DRM_RKGRUFA)

names(DRM_RKGRUFA)<- c("Participant", "Group", "Alcohol", "Related", "FalseAlarms")
DRM_RKGRUFA$Alcohol <- gl(2,53, labels = c("Before", "After"))
DRM_RKGRUFA$Group<-as.factor(DRM_RKGRUFA$Group)

View(DRM_RKGRUFA)
str(DRM_RKGRUFA)


##############################################################################################
#############################################################################################

#  False Alarms by related vs unrelated, between groups 

# Assess need for multilevel model

interceptdrmrufa <- gls(FalseAlarms~ 1, data = DRM_RKGRUFA, method = "ML")
randominterceptdrmrufa <- lme(FalseAlarms ~1, data= DRM_RKGRUFA, random = ~1|Group, method = "ML")
anova(interceptdrmrufa, randominterceptdrmrufa)

# Adding the random intercept did not improve the model.

# Set contrasts before building models. 

BeforevsAfter<-c(-1,1)
ControlvsMBO<-c(-1,1)
RelatedvsUnrelated <- c(-1,1)
contrasts(DRM_RKGRUFA$Alcohol)<-cbind(BeforevsAfter)
contrasts(DRM_RKGRUFA$Group)<-cbind(ControlvsMBO)
contrasts(DRM_RKGRUFA$Related)<-cbind(RelatedvsUnrelated)

# Check contrasts are correct

DRM_RKGRUFA$Alcohol
DRM_RKGRUFA$Group
DRM_RKGRUFA$Related

# Create baseline mixed effects model.  This is a Factorial Repeated-Measures Design

basedrmrufa <- lme(FalseAlarms ~ 1, random = ~1|Participant/Alcohol/Group/Related, 
                   data = DRM_RKGRUFA, method = "ML")   # This model predicts outcome by intercept, and the random part says that
                                                        # scores for each participant are within condition and group variables.

# Add main effects by updating base model.
drmrufa1 <- update(basedrmrufa, .~. + Group)  # Add group as a predictor of Hits
drmrufa2 <- update(drmrufa1, .~. +Alcohol)    # Add alcohol as a predictor of hits
drmrufa3 <- update(drmrufa2, .~. + Related)   # Add relatedness as a predictor of Hits
# Add interaction(s)
drmrufa4 <- update(drmrufa3, .~. +Group:Alcohol)
drmrufa5 <- update(drmrufa4, .~. + Group:Related)
drmrufa6 <- update(drmrufa5, .~. + Alcohol:Related)
drmrufa7 <- update(drmrufa6, .~. + Group:Alcohol:Related)

# Compare models
anova(basedrmrufa, drmrufa1, drmrufa2, drmrufa3, drmrufa4, drmrufa5, drmrufa6, drmrufa7)  # model drmrufa2, 3,6 are sig

summary(drmrufa6)
rcontrast(-15.228428,103) #effect size related/unrelated  "r =  0.832135875659438
rcontrast(3.096509,103) #effect size of interaction between alcohol and related/unrelated r = 0.291827076657492
rcontrast (-3.832929, 51)  # effect size of alcohol  r =  0.472908041033601

# Get descriptive statistics 
#I.Descriptives - Condition, all participants 
by(DRM_RKGRUFA$FalseAlarms, list(DRM_RKGRUFA$Related, DRM_RKGRUFA$Group,DRM_RKGRUFA$Alcohol ), stat.desc, basic=F)



############################################################################################################################################
######################   MBO Group RKG and FA's ##############################################

DRM_RKG_Within1 <- read_excel("~.....\\RKG_Within.xlsx")
DRM_RKG_Within <- as.data.frame(DRM_RKG_Within1)
str(DRM_RKG_Within)
DRM_RKG_Within$Participant <- as.numeric(as.integer(DRM_RKG_Within$Participant))

str(DRM_RKG_Within)
View(DRM_RKG_Within)

# Hits, within MBO group.  Totals as percentages, RKGs as proportions (%) of totals
DRM_RKGHitsMBO <- DRM_RKG_Within[1:69, c("Participant", "Alcohol", "HitTotal", "HitGuess", "HitKnow", "HitRemember")]
View(DRM_RKGHitsMBO)

# FA's, within MBO group.  Total FA's as %, RKG's as proportions of related and unrelated totals
DRM_RKGFAMBO <- DRM_RKG_Within[1:69, c("Participant","Alcohol", "FATotal", "RelFAGuess", "RelFAKnow", "RelFARemember", "RelFATotal",
                                       "UnrelFAGuess", "UnrelFAKnow", "UnrelFARemember", "UnrelFATotal")]

##############################  RKG by Hits, Within MBO Group  ###########################

# RKG Hits
DRM_RKGM <- melt.data.frame(DRM_RKGHitsMBO, id.vars = c("Participant", "Alcohol"),
                            measure.vars = c("HitGuess", "HitKnow", "HitRemember"))

View(DRM_RKGM)

# And for RKG hits data
names(DRM_RKGM)<- c("Participant", "Alcohol","RKG", "Hits")
str(DRM_RKGM)
DRM_RKGM$Alcohol<-as.factor(DRM_RKGM$Alcohol)
DRM_RKGM$RKG <- as.factor(DRM_RKGM$RKG)

str(DRM_RKGM)

############################################################################
#Set contrasts for RKG Hits
AftervsAfterMBO<-c(-1,1,0) 
BeforevsAfter<-c(1,0,-1)

contrasts(DRM_RKGM$Alcohol)<-cbind(AftervsAfterMBO, BeforevsAfter)

# Check contrasts are correct
DRM_RKGM$Alcohol


# Create baseline mixed effects model. 

basedrmrkghm <- lme(Hits ~ 1, random = ~1|Participant/Alcohol/RKG, 
                    data = DRM_RKGM, method = "ML", na.action = na.exclude)   
summary(basedrmrkghm)

# Add main effects by updating base model.
drmrkghm1 <- update(basedrmrkghm, .~. + Alcohol)  # Add alcohol as a predictor of MeanACC
drmrkghm2 <- update(drmrkghm1, .~. +RKG)    # Add RKG as a predictor
drmrkghm3 <- update(drmrkghm2, .~. +Alcohol:RKG) # Add interaction

anova(basedrmrkghm, drmrkghm1, drmrkghm2, drmrkghm3)

# Summary of drmrkghm2
summary(drmrkghm2)

#Bonferroni
pairwise.t.test(DRM_RKGM$Hits, DRM_RKGM$RKG, paired = TRUE, p.adjust.method = "bonferroni")
#          HitGuess HitKnow
#HitKnow     <2e-16   -      
#HitRemember <2e-16   0.0026

# Descriptives by condiion x alcohol
by(DRM_RKGM$Hits, list(DRM_RKGM$RKG), stat.desc, basic=F)


###################################################################################

############    MBO Group, FAs
# FA's, within MBO group.  Total FA's as %, RKG's as proportions of related and unrelated totals
DRM_RKGFAMBO <- DRM_RKG_Within[1:69, c("Participant","Alcohol", "FATotal", "RelFAGuess", "RelFAKnow", "RelFARemember", "RelFATotal",
                                       "UnrelFAGuess", "UnrelFAKnow", "UnrelFARemember", "UnrelFATotal")]

view(DRM_RKGFAMBO)

#FA's by Related / Unrelated
DRM_RUFAM <- melt.data.frame(DRM_RKGFAMBO, id.vars = c("Participant", "Alcohol"),
                             measure.vars = c("RelFATotal", "UnrelFATotal"))
view(DRM_RUFAM)

# For Related/Unrelated
names(DRM_RUFAM)<- c("Participant", "Alcohol", "RelUnrel", "FA")
str(DRM_RUFAM)
DRM_RUFAM$Alcohol<-as.factor(DRM_RUFAM$Alcohol)
str(DRM_RUFAM)

#########################################################################################
#########  False Alarms by Rel/Unrel, MBO Group #############

# Assess need for multilevel model

interceptdrmrufam <- gls(FA~ 1, data = DRM_RUFAM, method = "ML", na.action = na.exclude)
randominterceptdrmrufam <- lme(FA ~1, data= DRM_RUFAM, random = ~1|Participant/Alcohol, method = "ML", na.action = na.exclude)
anova(interceptdrmrufam, randominterceptdrmrufam)

# Adding the random intercept did not improve the model.

# Check variables are factors. 
str(DRM_RUFAM)
View(DRM_RUFAM)

# Set contrasts
BeforevsAfter<-c(1,0,-1)
BeforevsAfterMBO<- c(0,1,-1)
RelatedvsUnrelated <- c(-1,1)

contrasts(DRM_RUFAM$Alcohol)<-cbind(BeforevsAfter, BeforevsAfterMBO)
contrasts(DRM_RUFAM$RelUnrel) <-cbind(RelatedvsUnrelated)

# Check contrasts are correct

DRM_RUFAM$Alcohol
DRM_RUFAM$RelUnrel

# Create baseline mixed effects model

basedrmrufam <- lme(FA ~ 1, random = ~1|Participant/Alcohol/RelUnrel, 
                    data = DRM_RUFAM, method = "ML", na.action=na.exclude)

# Add main effects by updating base model.
drmrufa1m <- update(basedrmrufam, .~. + Alcohol)  # Add alcohol as a predictor of FAs
drmrufa2m <- update(drmrufa1m, .~. + RelUnrel) # Add related/unrelated as a predictor of FA's
drmrufa3m <- update(drmrufa2m, .~. +Alcohol:RelUnrel)# Add interaction

# Compare models
anova(basedrmrufam, drmrufa1m, drmrufa2m, drmrufa3m) # RelUnrel is sig, and interaction also sig.

summary(drmrufa3m)  # model 3 is highest level significant
rcontrast(-16.152660, 66) #r =  0.893369239904726
rcontrast(2.239186, 66) #r =  0.265716377486481 


####################################################################################################
##############    DRM - Sleep by DPrime   ######################

DRMCorr_Data1 <- read_excel("~.......\\DRM_Sleep.xlsx")
DRMCorr_Data <- as.data.frame(DRMCorr_Data1)
str(DRMCorr_Data)
DRM_RKG_Within$Participant <- as.numeric(as.integer(DRM_RKG_Within$Participant))
DRM_CorrDP <- DRMCorr_Data[1:23, c("Minutes", "dPrime")]
str(DRM_CorrDP)

modeldrmdp <- lm(formula = dPrime ~ Minutes, 
                 data = DRM_CorrDP)
summary(modeldrmdp)


baysianDRM = regressionBF(dPrime~.,data=DRM_CorrDP) 
baysianDRM


############################
########## DRM - Sleep by False Alarms

DRM_CorrFA <- DRMCorr_Data[1:23, c("Minutes", "TFA_T1T3")]
str(DRM_CorrFA)

modeldrmfa <- lm(formula = TFA_T1T3 ~ Minutes, 
                 data = DRM_CorrFA)
summary(modeldrmfa)

baysianDRMFA = regressionBF(TFA_T1T3 ~.,data=DRM_CorrFA) 
baysianDRMFA