##############################################################################
#                                                                            #        
#       Script for the "Measuring resilience prospectively                   #  
#           as the speed of affect recovery in daily life:                   #    
#         A complex systems perspective on mental health"                    #
#                               paper                                        #  
#                                                                            #  
##############################################################################

# Function to install a package or load if it exists. 
# Credits to: https://stackoverflow.com/a/9341833/5252007
load.package <- function(name) {
  if(!require(name, character.only = TRUE)) {
    install.packages(name, dep = TRUE)
    if(!require(name, character.only = TRUE)) stop("Package not found.")
  }
}

# Load the packages we will need.
load.package("readstata13") #For importing datasets
load.package("plyr")        #For data transformation, function "count"
load.package("dplyr")       #For data description, function "group_by" etc.
load.package("statar")      #For selecting groups
load.package("nlme")        #For multilevel models
load.package("DataCombine") #For laggind variables
load.package("qgraph")      #For network analysis
load.package("tidyr")       #To complete dataset
load.package("lattice")     #To plot missings
load.package("moments")     #For skewness and kurtosis 
load.package("robustHD")    #For standartization 

##Uploading the databases
my_data_main <- read.dta13("X:/My Desktop/Documents/twinssCan2.3.dta")

###Creating the subset with cross-sectional data
data_des <- subset(my_data_main, select = c("subjno","idtw", "_1_J1", "_1_J2", "_1_J3", "_1_J4", "_1_scl90_tot", "_2_scl90_tot", "gender", "age_wave1", "_2_le_neg"))

###Deleting parents and other unrelated participants (they do not have the age data): 

data_des<-data_des[!is.na(data_des$age_wave1), ]

##creating the variable representing the number of beeps for each individual: 
n_beeps <- plyr::count(data_des,c('subjno'))
data_des <- merge(n_beeps, data_des, by = "subjno")
rm(n_beeps)

####Sample selection process#### 
###Now the dataset is collapsed: 
data_des <- aggregate(.~subjno, data_des, mean, na.action=NULL) 

###Selection of the subsample with the lower level of happy childhood experiences: 

#checking the missings in the JTV items first: 

summary(data_des[c("_1_J1", "_1_J2", "_1_J3", "_1_J4")]) ## all items have the same number of missings 
## see the descriptive statistics part below for more details

data_des$subjno[is.na(data_des$`_1_J1`)] 
data_des$subjno[is.na(data_des$`_1_J2`)] 
data_des$subjno[is.na(data_des$`_1_J3`)] 
data_des$subjno[is.na(data_des$`_1_J4`)] ##and they belong to the same people

#Calculating then the JTV sum score: 

data_des$chi_exp <- data_des$`_1_J1` + data_des$`_1_J2` + data_des$`_1_J3` + data_des$`_1_J4` 

length(data_des$subjno[is.na(data_des$chi_exp)]) ## 25 people do not have values
data_des<-data_des[!is.na(data_des$chi_exp), ] ## they are omitted from the sample

#Splitting the sample to select the subsample with the lowest level of JTV sumscore (becasue items are formulating positively):  

data_des$chi_exp_2 <- xtile(data_des$chi_exp, n = 2)  

table(data_des$chi_exp_2) ## resulting in 451 people with lower level of JTV sumscore

data_des<-data_des[data_des$chi_exp_2 == 1, ] # Deleting people with high level of happy childhood


### SCL-90 scores: 

#Checking missings: 
summary(data_des[c("_1_scl90_tot", "_2_scl90_tot")]) ## see the descriptive statistics part below for more details

#Calculating the change score: 
data_des$scl_change <- data_des$`_2_scl90_tot` - data_des$`_1_scl90_tot` # thus people with missings either on T0 or T1 will have NAs 

length(data_des$subjno[!is.na(data_des$scl_change)]) # 249 participants have data availible 

# Deleting the participants without SCL scores: 

data_des<-data_des[!is.na(data_des$scl_change), ]

### Having more than 30% of ESM data available: 

length(data_des$subjno[(data_des$freq <= 18)]) #the study period consisted of 6 days with 10 beeps per day; 
                                               #60 is considered max (some people have more, but 60 was the aim)
                                               #and therefore 30% of 60 is 18. We exclude all people (10) with 
                                               #18 or less beeps:

data_des <- data_des[data_des$freq > 18, ] # resulting in 239 participants 

###Having at least one unpleasant experience 

events <- subset(my_data_main, select = c("subjno", "eve_pleasant"))
events <- events[events$subjno %in% data_des$subjno, ] #deleting extra people 
events$stress <- events$eve_pleasant <= 0
events <- aggregate(.~subjno, events, sum, na.action = na.omit) 
View(events) #"TWS0308" "TWS0609" - are these people; 
rm(events)

data_des <- data_des[!data_des$subjno == "TWS0308", ]
data_des <- data_des[!data_des$subjno == "TWS0609", ]


### Creating groups 

data_des$group <- xtile(data_des$scl_change, 3)
data_des$group <- factor(data_des$group, labels = c("Decrease group", "Stable group", "Increase group"))

table(data_des$group)

data_des <- data_des[!data_des$group == "Decrease group", ] #excluding people from the group with the decrease of symptoms

####Preparing variables for the analysis####

my_data <- subset(my_data_main, select = c("subjno", "dayno", "beepno", "idtw", "mood_cheerful", "mood_relaxed", "mood_insecur", "mood_lonely",  
                                             "mood_anxious", "mood_irritat", "mood_satisfi", "mood_listles", 
                                             "pat_suspic", "mood_down", "mood_guilty", "mood_enthus", 
                                             "phy_energy", "phy_tired", "eve_pleasant", "gender", "age_wave1"))
my_data <- my_data[my_data$subjno %in% data_des$subjno, ] #deleting extra people 
my_data <- merge(my_data, data_des[c("subjno", "group")], by = "subjno")


## NA factor and PA factor
my_data$NA_factor <- rowMeans(my_data[ , c("mood_insecur", "mood_lonely", "mood_anxious", "mood_irritat", "mood_listles", "pat_suspic", "mood_down", "mood_guilty")], na.rm=TRUE)
my_data$PA_factor <- rowMeans(my_data[ , c("mood_cheerful", "mood_relaxed","mood_satisfi",  "mood_enthus")], na.rm=TRUE)        

###Person-mean center these factors 
personmeans <- subset(my_data, select = c("subjno", "NA_factor", "PA_factor"))
personmeans <- aggregate(.~subjno, personmeans, mean, na.rm=TRUE)
names(personmeans)<-c("subjno", "NA_factor_mean", "PA_factor_mean")
my_data <-merge(my_data, personmeans, by="subjno")
my_data$NA_permean <- my_data$NA_factor - my_data$NA_factor_mean
my_data$PA_permean <- my_data$PA_factor - my_data$PA_factor_mean
rm(personmeans)

### Stressful event 

my_data$stress <- my_data$eve_pleasant * -1 
my_data$stress[my_data$stress < 0] <- NA

### We have a lot of missing time points, so need to fill in NAs there
my_data <- my_data[order(my_data$subjno, my_data$dayno, my_data$beepno),]

my_data_complete <- complete(my_data, nesting(group, idtw, subjno), dayno = 1:42, beepno=1:15) ### 42 is the maximum day number  

#Time variables, one for all days ##

my_data_complete$id_time <- rep(seq(1:630), 157)


##### Creating the lagged variable for unpleasant event ####

my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l1_stress", TimeVar = "beepno", slideBy =-1)
my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l2_stress", TimeVar = "beepno", slideBy =-2)
my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l3_stress", TimeVar = "beepno", slideBy =-3)
my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l4_stress", TimeVar = "beepno", slideBy =-4)

######Main analysis#########

###INVESTIGATING QUANTITATIVE DIFFERENCES BETWEEN GROUPS IN POSITIVE AND NEGATIVE AFFECT RECOVERY

## Negative affect 
#Lag 0
recovery_lag0_all_na <- lme(NA_permean ~ stress + id_time + group + stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag0_all_na)

#Assumptions: 

###Homogeniety of variance 
plot(recovery_lag0_all_na)
# Normality of residuals 
qqnorm(resid(recovery_lag0_all_na))
hist(resid(recovery_lag0_all_na))

#Lag 1
recovery_lag1_all_na <- lme(NA_permean ~  id_time + l1_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag1_all_na)

###Homogeniety of variance 
plot(jitter(fitted(recovery_lag1_all_na), amount=0.2), jitter(resid(recovery_lag1_all_na, type="p"), amount=0.3), pch=19, cex=.5, xlab="Fitted Values", ylab="Standardized Residuals")
# Normality of residuals 
qqnorm(resid(recovery_lag1_all_na))
hist(resid(recovery_lag1_all_na))

#Lag 2
recovery_lag2_all_na <- lme(NA_permean ~  id_time + l2_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag2_all_na)

###Homogeniety of variance 
plot(recovery_lag2_all_na)
# Normality of residuals 
qqnorm(resid(recovery_lag2_all_na))
hist(resid(recovery_lag2_all_na))

#Lag 3
recovery_lag3_all_na <- lme(NA_permean ~ id_time + l3_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l3_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag3_all_na)

###Homogeniety of variance 
plot(recovery_lag3_all_na)
# Normality of residuals 
qqnorm(resid(recovery_lag3_all_na))
hist(resid(recovery_lag3_all_na))

#Lag 4
recovery_lag4_all_na <- lme(NA_permean ~  id_time + l4_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l4_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag4_all_na)

###Homogeniety of variance 
plot(recovery_lag4_all_na)
# Normality of residuals 
qqnorm(resid(recovery_lag4_all_na))
hist(resid(recovery_lag4_all_na))

## Positive affect

#Lag 0
recovery_lag0_all_pa <- lme(PA_permean ~ stress + id_time + group + stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag0_all_pa)

###Homogeniety of variance 
plot(recovery_lag0_all_pa)
# Normality of residuals 
qqnorm(resid(recovery_lag0_all_pa))
hist(resid(recovery_lag0_all_pa))

###Homogeniety of variance 
plot(recovery_lag0_all_pa_rhythm)
# Normality of residuals 
qqnorm(resid(recovery_lag0_all_pa_rhythm))
hist(resid(recovery_lag0_all_pa_rhythm))

#Lag 1
recovery_lag1_all_pa <- lme(PA_permean ~  id_time + l1_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag1_all_pa) 

###Homogeniety of variance 
plot(recovery_lag1_all_pa)
# Normality of residuals 
qqnorm(resid(recovery_lag1_all_pa))
hist(resid(recovery_lag1_all_pa))

###Homogeniety of variance 
plot(recovery_lag1_all_pa_rhythm)
# Normality of residuals 
qqnorm(resid(recovery_lag1_all_pa_rhythm))
hist(resid(recovery_lag1_all_pa_rhythm))

#Lag 2
recovery_lag2_all_pa <- lme(PA_permean ~ l2_stress + id_time + group + l2_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag2_all_pa)

###Homogeniety of variance 
plot(recovery_lag2_all_pa)
# Normality of residuals 
qqnorm(resid(recovery_lag2_all_pa))
hist(resid(recovery_lag2_all_pa))

#Lag 3
recovery_lag3_all_pa <- lme(PA_permean ~ l3_stress + id_time + group + l3_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l3_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag3_all_pa)

###Homogeniety of variance 
plot(recovery_lag3_all_pa)
# Normality of residuals 
qqnorm(resid(recovery_lag3_all_pa))
hist(resid(recovery_lag3_all_pa))

#Lag 4
recovery_lag4_all_pa <- lme(PA_permean ~  id_time +l4_stress*group + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l4_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
summary(recovery_lag4_all_pa)

###Homogeniety of variance 
plot(recovery_lag4_all_pa)
# Normality of residuals 
qqnorm(resid(recovery_lag4_all_pa))
hist(resid(recovery_lag4_all_pa))

##IS THERE AN EFFECT FROM THE MEAN ON AN INDIVIDUAL GROUP LEVEL?####

### Lag 0
#NA 
#Stable group  
recovery_na_lag0_model_group1 <- lme(NA_permean ~ stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_na_lag0_model_group1)

###Homogeniety of variance 
plot(recovery_na_lag0_model_group1)
# Normality of residuals 
qqnorm(resid(recovery_na_lag0_model_group1))
hist(resid(recovery_na_lag0_model_group1))

#Increase group ###  
recovery_na_lag0_model_group2 <- lme(NA_permean ~ stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_na_lag0_model_group2) 

###Homogeniety of variance 
plot(recovery_na_lag0_model_group2)
# Normality of residuals 
qqnorm(resid(recovery_na_lag0_model_group2))
hist(resid(recovery_na_lag0_model_group2))

##PA
#Stable group ### 
recovery_na_lag0_model_group1 <- lme(PA_permean ~ stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_na_lag0_model_group1)

###Homogeniety of variance 
plot(recovery_na_lag0_model_group1)
# Normality of residuals 
qqnorm(resid(recovery_na_lag0_model_group1))
hist(resid(recovery_na_lag0_model_group1))

#Increase group ###
recovery_pa_lag0_model_group2 <- lme(PA_permean ~ stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_pa_lag0_model_group2)

###Homogeniety of variance 
plot(recovery_pa_lag0_model_group2)
# Normality of residuals 
qqnorm(resid(recovery_pa_lag0_model_group2))
hist(resid(recovery_pa_lag0_model_group2))

###Lag 1 

##NA 
#Stable group #
recovery_na_lag1_model_group1 <- lme(NA_permean ~ l1_stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_na_lag1_model_group1)

###Homogeniety of variance 
plot(recovery_na_lag1_model_group1)
# Normality of residuals 
qqnorm(resid(recovery_na_lag1_model_group1))
hist(resid(recovery_na_lag1_model_group1))

#Increase group #
recovery_na_lag1_model_group2 <- lme(NA_permean ~ l1_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_na_lag1_model_group2)

###Homogeniety of variance 
plot(recovery_na_lag1_model_group2)
# Normality of residuals 
qqnorm(resid(recovery_na_lag1_model_group2))
hist(resid(recovery_na_lag1_model_group2))

##PA
#Stable group #
recovery_pa_lag1_model_group1 <- lme(PA_permean ~ l1_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_pa_lag1_model_group1)

###Homogeniety of variance 
plot(recovery_pa_lag1_model_group1)
# Normality of residuals 
qqnorm(resid(recovery_pa_lag1_model_group1))
hist(resid(recovery_pa_lag1_model_group1))

#Increase group 
recovery_pa_lag1_model_group2 <- lme(PA_permean ~ l1_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_pa_lag1_model_group2)

###Homogeniety of variance 
plot(recovery_pa_lag1_model_group2)
# Normality of residuals 
qqnorm(resid(recovery_pa_lag1_model_group2))
hist(resid(recovery_pa_lag1_model_group2))

#Lag 2   
#NA
#Stable group
recovery_na_lag2_model_group1 <- lme(NA_permean ~ l2_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_na_lag2_model_group1) 

#Increase group
recovery_na_lag2_model_group2 <- lme(NA_permean ~ l2_stress+ id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_na_lag2_model_group2)

##PA
#Stable group
recovery_pa_lag2_model_group1 <- lme(PA_permean ~ l2_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_pa_lag2_model_group1)

#Increase group
recovery_pa_lag2_model_group2 <- lme(PA_permean ~ l2_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_pa_lag2_model_group2)


#Lag 3  
#NA
#table group
recovery_na_lag3_model_group1 <- lme(NA_permean ~ l3_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l3_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_na_lag3_model_group1) 

#Increase group
recovery_na_lag3_model_group2 <- lme(NA_permean ~ l3_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l3_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_na_lag3_model_group2)

##PA
#Stable group
recovery_pa_lag3_model_group1 <- lme(PA_permean ~ l3_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l3_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_pa_lag3_model_group1)

#Increase group
recovery_pa_lag3_model_group2 <- lme(PA_permean ~ l3_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l3_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_pa_lag3_model_group2)


#Lag 4  
#NA
#table group
recovery_na_lag4_model_group1 <- lme(NA_permean ~ l4_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l4_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_na_lag4_model_group1) 

#Increase group
recovery_na_lag4_model_group2 <- lme(NA_permean ~ l4_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l4_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_na_lag4_model_group2)

##PA
#Stable group
recovery_pa_lag4_model_group1 <- lme(PA_permean ~ l4_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l4_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Stable group")
summary(recovery_pa_lag4_model_group1)

#Increase group
recovery_pa_lag4_model_group2 <- lme(PA_permean ~ l4_stress + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l4_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset=group=="Increase group")
summary(recovery_pa_lag4_model_group2)



#####Investigating the effects of delayed recovery on the course of psychopathology ####

### NEGATIVE AFFECT
### Delayed recovery is operationalised as effect of random effect variances of the slope for stressor= BLUP ##
# We'll calculate AUC for T0, T1 and T2 
### At T0
recovery_na_all_t0 <- lme(NA_permean ~ stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
#summary(recovery_na_all_t0)

###Homogeniety of variance 
#plot(recovery_na_all_t1)
# Normality of residuals 
#qqnorm(resid(recovery_na_all_t1))
#hist(resid(recovery_na_all_t1))

### At T1
recovery_na_all_t1 <- lme(NA_permean ~ l1_stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
#summary(recovery_na_all_t1)

###Homogeniety of variance 
#plot(recovery_na_all_t1)
# Normality of residuals 
#qqnorm(resid(recovery_na_all_t1))
#hist(resid(recovery_na_all_t1))

### At T2
recovery_na_all_t2 <- lme(NA_permean ~ l2_stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
#summary(recovery_na_all_t2)

###Homogeniety of variance 
plot(recovery_na_all_t2)
# Normality of residuals 
#qqnorm(resid(recovery_na_all_t2))
#hist(resid(recovery_na_all_t2))

### Blups for T0 ##

na_blups_t0 <- as.data.frame(random.effects(recovery_na_all_t0, level = 2))
na_blups_t0$subjno <- rownames(random.effects(recovery_na_all_t0, level = 2))
na_blups_t0$subjno <- gsub('.*\\/', '', na_blups_t0$subjno)
na_blups_t0$idtw <- rownames(random.effects(recovery_na_all_t0, level = 2))
na_blups_t0$idtw <- sapply(strsplit(na_blups_t0$idtw, split='/', fixed=TRUE), function(x) (x[1]))
na_blups_t0$idtw <- as.numeric(na_blups_t0$idtw)
names(na_blups_t0)[3] <- "T0"
na_blups_t0$id_time <- NULL
na_blups_t0$`(Intercept)` <- NULL

### Blups for T1

na_blups_t1 <- as.data.frame(random.effects(recovery_na_all_t1, level = 2))
na_blups_t1$subjno <- rownames(random.effects(recovery_na_all_t1, level = 2))
na_blups_t1$subjno <- gsub('.*\\/', '', na_blups_t1$subjno)
na_blups_t1$idtw <- rownames(random.effects(recovery_na_all_t1, level = 2))
na_blups_t1$idtw <- sapply(strsplit(na_blups_t1$idtw, split='/', fixed=TRUE), function(x) (x[1]))
na_blups_t1$idtw <- as.numeric(na_blups_t1$idtw)
names(na_blups_t1)[3] <- "T1"
na_blups_t1$id_time <- NULL
na_blups_t1$`(Intercept)` <- NULL

### Blups for T2

na_blups_t2 <- as.data.frame(random.effects(recovery_na_all_t2, level = 2))
na_blups_t2$subjno <- rownames(random.effects(recovery_na_all_t2, level = 2))
na_blups_t2$subjno <- gsub('.*\\/', '', na_blups_t2$subjno)
na_blups_t2$idtw <- rownames(random.effects(recovery_na_all_t2, level = 2))
na_blups_t2$idtw <- sapply(strsplit(na_blups_t2$idtw, split='/', fixed=TRUE), function(x) (x[1]))
na_blups_t2$idtw <- as.numeric(na_blups_t2$idtw)
names(na_blups_t2)[3] <- "T2"
na_blups_t2$id_time <- NULL
na_blups_t2$`(Intercept)` <- NULL

### At this point there should be different number of people in Blups at lag 2 
## let's check 

### we lost two people; who are they? 
#setdiff(recovery_na_all_t1[["groups"]][["subjno"]], recovery_na_all_t2[["groups"]][["subjno"]])
# "4238/TWS0534" "4637/TWS0546"

na_blups_t0 <- na_blups_t0[!na_blups_t0$subjno == "TWS0534", ]
na_blups_t0 <- na_blups_t0[!na_blups_t0$subjno == "TWS0546", ]

na_blups_t1 <- na_blups_t1[!na_blups_t1$subjno == "TWS0534", ]
na_blups_t1 <- na_blups_t1[!na_blups_t1$subjno == "TWS0546", ]

###Now it should be possible to merge datasets; 

data_na_blups <- merge(na_blups_t0, na_blups_t1, by = c("subjno", "idtw")) 
data_na_blups <- merge(data_na_blups, na_blups_t2, by = c("subjno", "idtw")) 

###Calculating the AUC

data_na_blups$T0 <-  data_na_blups$T0 + as.numeric(fixed.effects(recovery_na_all_t0)[2])
data_na_blups$T1 <-  data_na_blups$T1 + as.numeric(fixed.effects(recovery_na_all_t1)[2])
data_na_blups$T2 <-  data_na_blups$T2 + as.numeric(fixed.effects(recovery_na_all_t2)[2])

data_na_blups$auc <- ((data_na_blups$T2 + data_na_blups$T1)/2 + (data_na_blups$T1 + data_na_blups$T0)/2) - 2*data_na_blups$T2
hist(data_na_blups$auc)

### Merge with dataset with other variables
data_na_blups <- merge(data_na_blups, data_des, by= c("subjno", "idtw")) 
names(data_na_blups)[names(data_na_blups) == '_1_scl90_tot'] <- 'SCL_T0'
names(data_na_blups)[names(data_na_blups) == '_2_scl90_tot'] <- 'SCL_T1'
names(data_na_blups)[names(data_na_blups) == '_2_le_neg'] <- 'neg_events'

###Standardisation: 
data_na_blups$SCL_T1_std <- standardize(data_na_blups$SCL_T1)
data_na_blups$auc_std <- standardize(data_na_blups$auc)
data_na_blups$scl_change_std <- standardize(data_na_blups$scl_change)


#### with covariances
blups_on_scl_na_std <- lme(SCL_T1_std ~ auc_std + gender + age_wave1 + neg_events + SCL_T0, random = list(idtw = ~ 1), data = data_na_blups, na.action=na.omit)
blups_on_scl_na_std <- lm(SCL_T1_std ~ auc_std + gender + age_wave1 + neg_events + SCL_T0, data = data_na_blups, na.action=na.omit)
summary(blups_on_scl_na_std)

### without covariances
blups_on_scl_na_nocov <- lme(SCL_T1 ~ auc + SCL_T0, random = list(idtw = ~ 1), data = data_na_blups, na.action=na.omit, control=list(opt="optim"))
summary(blups_on_scl_na_nocov)

##### Ploting

#ggplot(data_na_blups, aes(x=auc, y=scl_change, colour= as.factor(group))) + geom_point() + geom_smooth()
#ggplot(data_na_blups, aes(x=auc, y=scl_change)) + geom_point() + geom_smooth()
#ggplot(data_na_blups, aes(x=auc, y=scl_change)) + geom_point() + geom_smooth(method="lm", se=FALSE, colour="red")

ggplot(data_na_blups, aes(x=auc_std, y=scl_change_std)) + geom_point() + geom_smooth(method="lm", se=FALSE, colour="red") + labs(x = "AUC, standardized", y = "Change in SCL-90 scores, standardized", title = "Figure 3. Speed of negative affect recovery as a predictor of individual symptom trajectories")

####POSITIVE AFFECT##

### Delayed recovery is operationalised as effect of random effect variances of the slope for stressor= BLUP ##
# We'll calculate AUC for T0, T1 and T2 #

### At T0
recovery_pa_all_t0 <- lme(PA_permean ~ stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
#summary(recovery_pa_all_t0)

###Homogeniety of variance 
#plot(recovery_pa_all_t1)
# Normality of residuals 
#qqnorm(resid(recovery_pa_all_t1))
#hist(resid(recovery_pa_all_t1))

### At T1
recovery_pa_all_t1 <- lme(PA_permean ~ l1_stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
#summary(recovery_pa_all_t1)

###Homogeniety of variance 
#plot(recovery_pa_all_t1)
# Normality of residuals 
#qqnorm(resid(recovery_pa_all_t1))
#hist(resid(recovery_pa_all_t1))

### At T2
recovery_pa_all_t2 <- lme(PA_permean ~ l2_stress  + id_time + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l2_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
#summary(recovery_pa_all_t2)

###Homogeniety of variance 
#plot(recovery_pa_all_t2)
# Normality of residuals 
#qqnorm(resid(recovery_pa_all_t2))
#hist(resid(recovery_pa_all_t2))

### Blups for T0

pa_blups_t0 <- as.data.frame(random.effects(recovery_pa_all_t0, level = 2))
pa_blups_t0$subjno <- rownames(random.effects(recovery_pa_all_t0, level = 2))
pa_blups_t0$subjno <- gsub('.*\\/', '', pa_blups_t0$subjno)
pa_blups_t0$idtw <- rownames(random.effects(recovery_pa_all_t0, level = 2))
pa_blups_t0$idtw <- sapply(strsplit(pa_blups_t0$idtw, split='/', fixed=TRUE), function(x) (x[1]))
pa_blups_t0$idtw <- as.numeric(pa_blups_t0$idtw)
names(pa_blups_t0)[3] <- "T0"
pa_blups_t0$id_time <- NULL
pa_blups_t0$`(Intercept)` <- NULL

### Blups for T1

pa_blups_t1 <- as.data.frame(random.effects(recovery_pa_all_t1, level = 2))
pa_blups_t1$subjno <- rownames(random.effects(recovery_pa_all_t1, level = 2))
pa_blups_t1$subjno <- gsub('.*\\/', '', pa_blups_t1$subjno)
pa_blups_t1$idtw <- rownames(random.effects(recovery_pa_all_t1, level = 2))
pa_blups_t1$idtw <- sapply(strsplit(pa_blups_t1$idtw, split='/', fixed=TRUE), function(x) (x[1]))
pa_blups_t1$idtw <- as.numeric(pa_blups_t1$idtw)
names(pa_blups_t1)[3] <- "T1"
pa_blups_t1$id_time <- NULL
pa_blups_t1$`(Intercept)` <- NULL

### Blups for T2

pa_blups_t2 <- as.data.frame(random.effects(recovery_pa_all_t2, level = 2))
pa_blups_t2$subjno <- rownames(random.effects(recovery_pa_all_t2, level = 2))
pa_blups_t2$subjno <- gsub('.*\\/', '', pa_blups_t2$subjno)
pa_blups_t2$idtw <- rownames(random.effects(recovery_pa_all_t2, level = 2))
pa_blups_t2$idtw <- sapply(strsplit(pa_blups_t2$idtw, split='/', fixed=TRUE), function(x) (x[1]))
pa_blups_t2$idtw <- as.numeric(pa_blups_t2$idtw)
names(pa_blups_t2)[3] <- "T2"
pa_blups_t2$id_time <- NULL
pa_blups_t2$`(Intercept)` <- NULL


### At this point there should be different number of people in Blups at lag 2 

pa_blups_t0 <- pa_blups_t0[!pa_blups_t0$subjno == "TWS0534", ]
pa_blups_t0 <- pa_blups_t0[!pa_blups_t0$subjno == "TWS0546", ]

pa_blups_t1 <- pa_blups_t1[!pa_blups_t1$subjno == "TWS0534", ]
pa_blups_t1 <- pa_blups_t1[!pa_blups_t1$subjno == "TWS0546", ]

###Now it should be possible to merge datasets; 
data_pa_blups <- merge(pa_blups_t0, pa_blups_t1, by = c("subjno", "idtw")) 
data_pa_blups <- merge(data_pa_blups, pa_blups_t2, by = c("subjno", "idtw")) 

###Calculating the AUC

data_pa_blups$T0 <-  data_pa_blups$T0 + as.numeric(fixed.effects(recovery_pa_all_t0)[2])
data_pa_blups$T1 <-  data_pa_blups$T1 + as.numeric(fixed.effects(recovery_pa_all_t0)[2])
data_pa_blups$T2 <-  data_pa_blups$T2 + as.numeric(fixed.effects(recovery_pa_all_t0)[2])

data_pa_blups$auc <- ((data_pa_blups$T2 + data_pa_blups$T1)/2 + (data_pa_blups$T1 + data_pa_blups$T0)/2) - 2*data_pa_blups$T2
hist(data_pa_blups$auc)


### Merge with dataset with other variables
data_pa_blups <- merge(data_pa_blups, data_des, by= c("subjno", "idtw")) 
names(data_pa_blups)[names(data_pa_blups) == '_1_scl90_tot'] <- 'SCL_T0'
names(data_pa_blups)[names(data_pa_blups) == '_2_scl90_tot'] <- 'SCL_T1'
names(data_pa_blups)[names(data_pa_blups) == '_2_le_neg'] <- 'neg_events'

###Standardisation: 
data_pa_blups$SCL_T1_std <- standardize(data_pa_blups$SCL_T1)
data_pa_blups$auc_std <- standardize(data_pa_blups$auc)
data_pa_blups$scl_change_std <- standardize(data_pa_blups$scl_change)

#### with covariances
blups_on_scl_pa_std <- lme(SCL_T1_std ~ auc_std + gender + age_wave1 + neg_events + SCL_T0, random = list(idtw = ~ 1), data = data_pa_blups, na.action=na.omit)
blups_on_scl_pa_std <- lm(SCL_T1_std ~ auc_std + gender + age_wave1 + neg_events + SCL_T0, data = data_pa_blups, na.action=na.omit)
summary(blups_on_scl_pa_std)


### without covariances
blups_on_scl_pa_nocov <- lme(SCL_T1 ~ auc + SCL_T0, random = list(idtw = ~ 1), data = data_pa_blups, na.action=na.omit, control=list(opt="optim"))
summary(blups_on_scl_pa_nocov)


##### Ploting

#ggplot(data_pa_blups, aes(x=auc, y=scl_change, colour= as.factor(group))) + geom_point() + geom_smooth()
#ggplot(data_pa_blups, aes(x=auc, y=scl_change)) + geom_point() + geom_smooth()
ggplot(data_pa_blups, aes(x=auc_std, y=scl_change_std)) + geom_point() + geom_smooth(method="lm", se=FALSE, colour="red")


##### Descriptive analysis ####


#Number of people# 
table(data_des$group)

#Gender#
prop.table(table(data_des$gender[data_des$group == "Stable group"], exclude = NULL))
prop.table(table(data_des$gender[data_des$group == "Increase group"], exclude = NULL))
prop.table(table(data_des$gender[data_des$group == "Decrease group"], exclude = NULL))

#And is it different between groups? 
chisq.test(data_des$gender, data_des$group) #No

##Eductation and Ethnicity
##These data come from additional dataset with demographic variables: 
my_data_demo <- read.dta13("X:/My Desktop/Documents/demografie.dta") 
names(my_data_demo)
data_demo <- subset(my_data_demo, select = c("st_subjid", "educ_cb3", "etn_b","etn_z", "etn_a", "etn_r", "etn2"))
data_demo$subjno <- data_demo$st_subjid # because subjno has a different name in this dataset
data_demo$st_subjid <- NULL
data_des <- merge(data_demo, data_des,  by = "subjno")
rm(data_demo)

# Education: 
table(data_des$educ_cb3[data_des$group == "Stable group"], exclude = NULL)
table(data_des$educ_cb3[data_des$group == "Increase group"], exclude = NULL)

#difference between groups: 
chisq.test(data_des$educ_cb3, data_des$group) #No

#Ethnicity: 

table(data_des$etn2, data_des$group)
table(data_des$etn_b, data_des$group)
table(data_des$etn_z, data_des$group)
table(data_des$etn_a, data_des$group)
table(data_des$etn_r, data_des$group)

#Age 
options(pillar.sigfig=4) ## Becuase dplyr "ëats" digits, apparently 

group_by(data_des, group) %>%
  summarise(
    count = n(),
    mean = mean(age_wave1, na.rm = TRUE),
    min = min(age_wave1, na.rm = TRUE),
    max = max(age_wave1, na.rm = TRUE),
    sd = sd(age_wave1, na.rm = TRUE)
  )

#Is it different? 
var.test(data_des$age_wave1 ~ data_des$group) # variances are similar
t.test(data_des$age_wave1 ~ data_des$group, var.equal = T) ## No differences in age

#JTV scores
group_by(data_des, group) %>%
  summarise(
    count = n(),
    mean = mean(chi_exp, na.rm = TRUE),
    min = min(chi_exp, na.rm = TRUE),
    max = max(chi_exp, na.rm = TRUE),
    sd = sd(chi_exp, na.rm = TRUE)
  )

var.test(data_des$chi_exp ~ data_des$group) # variances are different
t.test(data_des$chi_exp ~ data_des$group, var.equal = F) ## Different level, smaller in the Increase group

#SCL-90 at baseline
group_by(data_des, group) %>%
  summarise(
    count = n(),
    mean = mean(`_1_scl90_tot` + 90, na.rm = TRUE),
    min = min(`_1_scl90_tot` + 90, na.rm = TRUE),
    max = max(`_1_scl90_tot` + 90, na.rm = TRUE),
    sd = sd(`_1_scl90_tot`, na.rm = TRUE)
  )

var.test(data_des$`_1_scl90_tot` ~ data_des$group) # variances are different
t.test(data_des$`_1_scl90_tot` ~ data_des$group, var.equal = F) ## Not different

##comparison with the Decrease group
#TukeyHSD(aov(data_des$`_1_scl90_tot` ~ data_des$group))

#SCL-90 at follow-up
group_by(data_des, group) %>%
  summarise(
    count = n(),
    mean = mean(`_2_scl90_tot` + 90, na.rm = TRUE),
    min = min(`_2_scl90_tot` + 90, na.rm = TRUE),
    max = max(`_2_scl90_tot` + 90, na.rm = TRUE),
    sd = sd(`_2_scl90_tot` + 90, na.rm = TRUE)
  )

var.test(data_des$`_2_scl90_tot` ~ data_des$group) # variances are different
t.test(data_des$`_2_scl90_tot` ~ data_des$group, var.equal = F) ## Different

#SCL change 
group_by(data_des, group) %>%
  summarise(
    count = n(),
    mean = mean(scl_change, na.rm = TRUE),
    min = min(scl_change, na.rm = TRUE),
    max = max(scl_change, na.rm = TRUE),
    sd = sd(scl_change, na.rm = TRUE)
  )


##Negative life events between T0 and T1
names(data_des)[names(data_des) == '_2_le_neg'] <- 'neg_events'

group_by(data_des, group) %>%
  summarise(
    count = n(),
    mean = mean(neg_events, na.rm = TRUE),
    min = min(neg_events, na.rm = TRUE),
    max = max(neg_events, na.rm = TRUE),
    sd = sd(neg_events, na.rm = TRUE)
  )

var.test(data_des$neg_events ~ data_des$group) # variances are different
t.test(data_des$neg_events ~ data_des$group, var.equal = F) ## same

###  ESM variables ###

###Number of stressful events: 
events <- merge(events, data_des[c("subjno", "group")], by = "subjno")

group_by(events, group) %>%
  summarise(
    count = n(),
    mean = mean(stress, na.rm = TRUE),
    min = min(stress, na.rm = TRUE),
    max = max(stress, na.rm = TRUE),
    sd = sd(stress, na.rm = TRUE)
  )

var.test(events$stress ~ data_des$group) # variances are equal
t.test(events$stress ~ data_des$group, var.equal = T) ## same


##Get ESM variabels means and SDs; 
# Because we need within-person SD, we use this function (written by Robert Walker, found on stackoverflow.com) : 

XTSUM <- function(data, varname, unit) {
  varname <- enquo(varname)
  loc.unit <- enquo(unit)
  ores <- data %>% summarise(ovr.mean=mean(!! varname, na.rm=TRUE), ovr.sd=sd(!! varname, na.rm=TRUE), ovr.min = min(!! varname, na.rm=TRUE), ovr.max=max(!! varname, na.rm=TRUE), ovr.N=sum(as.numeric((!is.na(!! varname)))))
  bmeans <- data %>% group_by(!! loc.unit) %>% summarise(meanx=mean(!! varname, na.rm=T), t.count=sum(as.numeric(!is.na(!! varname))))
  bres <- bmeans %>% ungroup() %>% summarise(between.sd = sd(meanx, na.rm=TRUE), between.min = min(meanx, na.rm=TRUE), between.max=max(meanx, na.rm=TRUE), Units=sum(as.numeric(!is.na(t.count))), t.bar=mean(t.count, na.rm=TRUE))
  wdat <- data %>% group_by(!! loc.unit) %>% mutate(W.x = scale(!! varname, scale=FALSE))
  wres <- wdat %>% ungroup() %>% summarise(within.sd=sd(W.x, na.rm=TRUE), within.min=min(W.x, na.rm=TRUE), within.max=max(W.x, na.rm=TRUE))
  return(list(ores=ores,bres=bres,wres=wres))
}

XTSUM(subset(my_data_complete, group == 'Stable group'), varname=NA_factor, unit=subjno)
XTSUM(subset(my_data_complete, group == 'Stable group'), varname=PA_factor, unit=subjno)
XTSUM(subset(my_data_complete, group == 'Stable group'), varname=stress, unit=subjno)

XTSUM(subset(my_data_complete, group == 'Increase group'), varname=NA_factor, unit=subjno)
XTSUM(subset(my_data_complete, group == 'Increase group'), varname=PA_factor, unit=subjno)
XTSUM(subset(my_data_complete, group == 'Increase group'), varname=stress, unit=subjno)


###ESM variables; for them to get means and SDs dataset must be aggregated, but to compare groups, 

# multilevel models must be used (not nessesary, but in this way more data is saved)

## comparing NA factor between groups

Na_factor_T0_difference <- lme(NA_factor ~ group, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim")) 
summary(Na_factor_T0_difference) ##NO
rm(Na_factor_T0_difference)

##PA affect score

PA_factor_T0_difference <- lme(PA_factor ~ group, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))

summary(PA_factor_T0_difference) ##NO
rm(PA_factor_T0_difference)

##Daily stressors mean 

# Are they different between groups? 
stress_group_difference <- lme(stress ~ group, data = my_data,  random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))
summary(stress_group_difference)### No difference
rm(stress_group_difference)

###Daily stressors sequences: are they same across groups? 

my_data_complete$cons_stress <- my_data_complete$stress &  my_data_complete$l1_stress >= 0
table(my_data_complete$cons_stress)
summary(lme(cons_stress ~ group12, data = my_data_complete,  random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim")))

my_data_complete$cons_stress2 <- my_data_complete$stress &  my_data_complete$l1_stress & my_data_complete$l2_stress >= 0
summary(lme(cons_stress2 ~ group12, data = my_data_complete,  random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim")))


### how many missings there are? 

### it is impossible to work with complete dataset, because 42 mostly empty rows add too much noise. Therefore, up to 8 days for this analysis

my_data_missings <- complete(my_data, nesting(group, idtw, subjno), dayno = 1:8, beepno=1:10) ###   

#Time variables, one for all days ##

my_data_missings$morning <- my_data_missings$beepno < 4
my_data_missings$afternoon <- my_data_missings$beepno >= 4 & my_data_missings$beepno <= 7  
my_data_missings$evening <- my_data_missings$beepno > 7

my_data_missings$count <- is.na(my_data_missings$NA_factor)

table(my_data_missings$count, my_data_missings$group)
chisq.test(my_data_missings$count, my_data_missings$group, correct=FALSE)

###Easier alterpative
my_data_missings$time_of_the_day <- factor(cut(my_data_missings$beepno, breaks=c(0,3,7,11), ordered=TRUE, labels = c("Morning", "Afternoon", "Evening")))
chisq.test(my_data_missings$count, my_data_missings$group, my_data_missings$time_of_the_day, correct=FALSE) ## Something is different:

prop.table(table(my_data_missings$count[my_data_missings$morning == T], my_data_missings$group[my_data_missings$morning == T]))
chisq.test(my_data_missings$count[my_data_missings$morning == T], my_data_missings$group[my_data_missings$morning == T], correct = F)

prop.table(table(my_data_missings$count[my_data_missings$afternoon == T], my_data_missings$group[my_data_missings$afternoon == T]))
chisq.test(my_data_missings$count[my_data_missings$afternoon == T], my_data_missings$group[my_data_missings$afternoon == T], correct = F)

prop.table(table(my_data_missings$count[my_data_missings$evening == T], my_data_missings$group[my_data_missings$evening == T]))
chisq.test(my_data_missings$count[my_data_missings$evening == T], my_data_missings$group[my_data_missings$evening == T], correct = F) ### Apperantly differences are in the evening


table(my_data_missings$count, my_data_missings$group, my_data_missings$dayno)

barchart(table(my_data_missings$dayno,  my_data_missings$group, my_data_missings$count), origin = 0, auto.key=T, ylab = "Day number")

barchart(table(my_data_missings$beepno,  my_data_missings$group, my_data_missings$count), origin = 0, auto.key=T)
barchart(table(my_data_missings$time_of_the_day,  my_data_missings$group, my_data_missings$count), origin = 0, auto.key=T, ylab = "Time of the day")

chisq.test(my_data_missings$time_of_the_day,  my_data_missings$group, my_data_missings$count)

#### Check whether SDs of NA and PA and stressors are the same between groups

test_NA <- lme(NA_factor ~ group + id_time, data = my_data_complete, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))
test_NA2 <- lme(NA_factor ~ group + id_time, data = my_data_complete, random=~1|idtw/subjno, weights=varIdent(form=~1|group), na.action=na.omit, control=list(opt="optim"))

anova(test_NA, test_NA2)

test_PA <- lme(PA_factor ~ group + id_time, data = my_data_complete, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))
test_PA2 <- lme(PA_factor ~ group + id_time, data = my_data_complete, random=~1|idtw/subjno, weights=varIdent(form=~1|group), na.action=na.omit, control=list(opt="optim"))

anova(test_PA, test_PA2)

test_stress <- lme(stress ~ group + id_time, data = my_data_complete, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))
test_stress2 <- lme(stress ~ group + id_time, data = my_data_complete, random=~1|idtw/subjno, weights=varIdent(form=~1|group), na.action=na.omit, control=list(opt="optim"))

anova(test_stress, test_stress2) ### this one IS the same, good 


###Multiverse analysis#####

# Assumptions: maximum power
# Same level of symptoms on baseline, different at T1
# Same level of affect and stressors 

rm(data_des)

###Creating the subset with cross-sectional data
data_des <- subset(my_data_main, select = c("subjno","idtw", "_1_J1", "_1_J2", "_1_J3", "_1_J4", "_1_scl90_tot", "_2_scl90_tot", "gender", "age_wave1", "_2_le_neg"))

###Deleting parents and other unrelated participants (they do not have the age data): 

data_des<-data_des[!is.na(data_des$age_wave1), ]

##creating the variable representing the number of beeps for each individual: 
n_beeps <- plyr::count(data_des,c('subjno'))
data_des <- merge(n_beeps, data_des, by = "subjno")
rm(n_beeps)

###Now the dataset is collapsed: 
data_des <- aggregate(.~subjno, data_des, mean, na.action=NULL) 

###Selection of the subsample with the lower level of happy childhood experiences: 

#checking the missings in the JTV items first: 

summary(data_des[c("_1_J1", "_1_J2", "_1_J3", "_1_J4")]) ## all items have the same number of missings 
## see the descriptive statistics part below for more details

data_des$subjno[is.na(data_des$`_1_J1`)] 
data_des$subjno[is.na(data_des$`_1_J2`)] 
data_des$subjno[is.na(data_des$`_1_J3`)] 
data_des$subjno[is.na(data_des$`_1_J4`)] ##and they belong to the same people

#Calculating then the JTV sum score: 

data_des$chi_exp <- data_des$`_1_J1` + data_des$`_1_J2` + data_des$`_1_J3` + data_des$`_1_J4` 

length(data_des$subjno[is.na(data_des$chi_exp)]) ## 25 people do not have values
data_des<-data_des[!is.na(data_des$chi_exp), ] ## they are omitted from the sample

#Splitting the sample to select the subsample with the lowest level of JTV sumscore (becasue items are formulating positively):  

data_des$chi_exp_2 <- xtile(data_des$chi_exp, n = 2)  

table(data_des$chi_exp_2) ## resulting in 451 people with lower level of JTV sumscore

data_des<-data_des[data_des$chi_exp_2 == 1, ] # Deleting people with high level of happy childhood


### SCL-90 scores: 

#Checking missings: 
summary(data_des[c("_1_scl90_tot", "_2_scl90_tot")]) ## see the descriptive statistics part below for more details

#Calculating the change score: 
data_des$scl_change <- data_des$`_2_scl90_tot` - data_des$`_1_scl90_tot` # thus people with missings either on T0 or T1 will have NAs 

length(data_des$subjno[!is.na(data_des$scl_change)]) # 249 participants have data availible 

# Deleting the participants without SCL scores: 

data_des<-data_des[!is.na(data_des$scl_change), ]

### Having more than 30% of ESM data available: 

length(data_des$subjno[(data_des$freq <= 18)]) #the study period consisted of 6 days with 10 beeps per day; 
#60 is considered max (some people have more, but 60 was the aim)
#and therefore 30% of 60 is 18. We exclude all people (10) with 
#18 or less beeps:

data_des <- data_des[data_des$freq > 18, ] # resulting in 239 participants 

###Having at least one unpleasant experience 

events <- subset(my_data_main, select = c("subjno", "eve_pleasant"))
events <- events[events$subjno %in% data_des$subjno, ] #deleting extra people 
events$stress <- events$eve_pleasant <= 0
events <- aggregate(.~subjno, events, sum, na.action = na.omit) 
View(events) #"TWS0308" "TWS0609" - are these people; 
rm(events)

data_des <- data_des[!data_des$subjno == "TWS0308", ]
data_des <- data_des[!data_des$subjno == "TWS0609", ]

####Group allocations:
# To make it reproducible, we shall exclude the same Decrease group, so: 

data_des$group <- xtile(data_des$scl_change, 3)
data_des$group <- factor(data_des$group, labels = c("Decrease group", "Stable group", "Increase group"))


data_des <- data_des[!data_des$group == "Decrease group", ] #excluding people from the group with the decrease of symptoms
data_des$group = NULL
##


# we need at least 70 people in each group.
data_des$group1 <-xtile(data_des$scl_change, cutpoints = c(1, 6)) # these are limits upper and lower

xtile(data_des$scl_change, cutpoints = -11) # -13, -12, -11 - only versions "from the bottom"

table(xtile(data_des$scl_change, cutpoints = 7)) #1, 2, 3, 4 (the actual), 5, 6

table(xtile(data_des$scl_change, cutpoints = c(1, 2))) #two - 1 and 6, 2 and 6, 3 and 6, 4 and 6, 5 and 6, 1 and 5, 2 and 5, 3 and 5
# 4 and 5, 1 and 4, 2 and 4, 3 and 4, 1 and 3, 2 and 3, 1 and 2
table(xtile(data_des$scl_change, cutpoints = c(-13, 2)))# -13 $ 2, 3, 4, 5, 6
table(xtile(data_des$scl_change, cutpoints = c(-12, 7)))# -12 $ 5, 6
table(xtile(data_des$scl_change, cutpoints = c(-11, 6)))# -11  6



#creating all possible groupings
# without negative "tail"
# 8 groups 
data_des$group1 <- xtile(data_des$scl_change, cutpoints = c(-13, 2)) # we need group 2 and 3
data_des$group2 <- xtile(data_des$scl_change, cutpoints = c(-13, 3))
data_des$group3 <- xtile(data_des$scl_change, cutpoints = c(-13, 4))
data_des$group4 <- xtile(data_des$scl_change, cutpoints = c(-13, 5))
data_des$group5 <- xtile(data_des$scl_change, cutpoints = c(-13, 6))
data_des$group6 <- xtile(data_des$scl_change, cutpoints = c(-12, 5))
data_des$group7 <- xtile(data_des$scl_change, cutpoints = c(-12, 6))
data_des$group8 <- xtile(data_des$scl_change, cutpoints = c(-11, 6))

#With only one split (maximum power!)
data_des$group9 <- xtile(data_des$scl_change, cutpoints = 1)
data_des$group10 <- xtile(data_des$scl_change, cutpoints = 2)
data_des$group11 <- xtile(data_des$scl_change, cutpoints = 3)
data_des$group12 <- xtile(data_des$scl_change, cutpoints = 4) #the actual grouping
data_des$group13 <- xtile(data_des$scl_change, cutpoints = 5)
data_des$group14 <- xtile(data_des$scl_change, cutpoints = 6)

## With potential group in between
data_des$group15 <- xtile(data_des$scl_change, cutpoints = c(1, 6))
data_des$group16 <- xtile(data_des$scl_change, cutpoints = c(2, 6))
data_des$group17 <- xtile(data_des$scl_change, cutpoints = c(3, 6))
data_des$group18 <- xtile(data_des$scl_change, cutpoints = c(4, 6))
data_des$group19 <- xtile(data_des$scl_change, cutpoints = c(5, 6))
data_des$group20 <- xtile(data_des$scl_change, cutpoints = c(1, 5))
data_des$group21 <- xtile(data_des$scl_change, cutpoints = c(2, 5))
data_des$group22 <- xtile(data_des$scl_change, cutpoints = c(3, 5))
data_des$group23 <- xtile(data_des$scl_change, cutpoints = c(4, 5))
data_des$group24 <- xtile(data_des$scl_change, cutpoints = c(1, 4))
data_des$group25 <- xtile(data_des$scl_change, cutpoints = c(2, 4))
data_des$group26 <- xtile(data_des$scl_change, cutpoints = c(3, 4))
data_des$group27 <- xtile(data_des$scl_change, cutpoints = c(1, 3))
data_des$group28 <- xtile(data_des$scl_change, cutpoints = c(2, 3))
data_des$group29 <- xtile(data_des$scl_change, cutpoints = c(1, 2))

###Checking SCL-90 at baseline: 

summary(lm(data_des$`_1_scl90_tot` ~ data_des$group1, subset = data_des$group1 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group2, subset = data_des$group2 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group3, subset = data_des$group3 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group4, subset = data_des$group4 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group5, subset = data_des$group5 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group6, subset = data_des$group6 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group7, subset = data_des$group7 != 1))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group8, subset = data_des$group8 != 1))

summary(lm(data_des$`_1_scl90_tot` ~ data_des$group9))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group10))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group11))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group12))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group13))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group14))

summary(lm(data_des$`_1_scl90_tot` ~ data_des$group15, subset = data_des$group15 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group16, subset = data_des$group16 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group17, subset = data_des$group17 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group18, subset = data_des$group18 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group19, subset = data_des$group19 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group20, subset = data_des$group20 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group21, subset = data_des$group21 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group22, subset = data_des$group22 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group23, subset = data_des$group23 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group24, subset = data_des$group24 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group25, subset = data_des$group25 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group26, subset = data_des$group26 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group27, subset = data_des$group27 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group28, subset = data_des$group28 != 2))
summary(lm(data_des$`_1_scl90_tot` ~ data_des$group28, subset = data_des$group29 != 2))

###Checking SCL-90 at T1: 

summary(lm(data_des$`_2_scl90_tot` ~ data_des$group1, subset = data_des$group1 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group2, subset = data_des$group2 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group3, subset = data_des$group3 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group4, subset = data_des$group4 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group5, subset = data_des$group5 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group6, subset = data_des$group6 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group7, subset = data_des$group7 != 1))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group8, subset = data_des$group8 != 1))

summary(lm(data_des$`_2_scl90_tot` ~ data_des$group9))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group10))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group11))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group12))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group13))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group14))

summary(lm(data_des$`_2_scl90_tot` ~ data_des$group15, subset = data_des$group15 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group16, subset = data_des$group16 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group17, subset = data_des$group17 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group18, subset = data_des$group18 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group19, subset = data_des$group19 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group20, subset = data_des$group20 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group21, subset = data_des$group21 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group22, subset = data_des$group22 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group23, subset = data_des$group23 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group24, subset = data_des$group24 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group25, subset = data_des$group25 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group26, subset = data_des$group26 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group27, subset = data_des$group27 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group28, subset = data_des$group28 != 2))
summary(lm(data_des$`_2_scl90_tot` ~ data_des$group28, subset = data_des$group29 != 2))

####Preparing variables for the analysis

my_data <- subset(my_data_main, select = c("subjno", "dayno", "beepno", "mood_cheerful", "mood_relaxed", "mood_insecur", "mood_lonely",  
                                           "mood_anxious", "mood_irritat", "mood_satisfi", "mood_listles", 
                                           "pat_suspic", "mood_down", "mood_guilty", "mood_enthus", 
                                           "phy_energy", "phy_tired", "eve_pleasant"))
my_data <- my_data[my_data$subjno %in% data_des$subjno, ] #deleting extra people 
my_data <- merge(my_data, data_des, by = "subjno")


## NA factor and PA factor
my_data$NA_factor <- rowMeans(my_data[ , c("mood_insecur", "mood_lonely", "mood_anxious", "mood_irritat", "mood_listles", "pat_suspic", "mood_down", "mood_guilty")], na.rm=TRUE)
my_data$PA_factor <- rowMeans(my_data[ , c("mood_cheerful", "mood_relaxed","mood_satisfi",  "mood_enthus")], na.rm=TRUE)        

###Person-mean center these factors 
personmeans <- subset(my_data, select = c("subjno", "NA_factor", "PA_factor"))
personmeans <- aggregate(.~subjno, personmeans, mean, na.rm=TRUE)
names(personmeans)<-c("subjno", "NA_factor_mean", "PA_factor_mean")
my_data <-merge(my_data, personmeans, by="subjno")
my_data$NA_permean <- my_data$NA_factor - my_data$NA_factor_mean
my_data$PA_permean <- my_data$PA_factor - my_data$PA_factor_mean
rm(personmeans)

### Stressful event 

my_data$stress <- my_data$eve_pleasant * -1 
my_data$stress[my_data$stress < 0] <- NA

### We have a lot of missing time points, so need to fill in NAs there
my_data <- my_data[order(my_data$subjno, my_data$dayno, my_data$beepno),]

my_data_complete <- complete(my_data, nesting(idtw, subjno), dayno = 1:42, beepno=1:15) ### 42 is the maximum day number  

#Time variables, one for all days ##

my_data_complete$id_time <- rep(seq(1:630), 157)


##### Creating the lagged variable for unpleasant event 

my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l1_stress", TimeVar = "beepno", slideBy =-1)
my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l2_stress", TimeVar = "beepno", slideBy =-2)
my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l3_stress", TimeVar = "beepno", slideBy =-3)
my_data_complete = slide(data = my_data_complete, Var = "stress", GroupVar = "subjno", NewVar = "l4_stress", TimeVar = "beepno", slideBy =-4)

#### Checking the difference in NA and stressors: 

summary(lme(NA_factor ~ group1, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group2, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group3, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group4, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group5, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group6, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group7, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(NA_factor ~ group8, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 

summary(lme(NA_factor ~ group9, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(NA_factor ~ group10, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(NA_factor ~ group11, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(NA_factor ~ group12, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(NA_factor ~ group13, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(NA_factor ~ group14, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 

summary(lme(NA_factor ~ group15, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group16, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group17, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group18, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group19, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group20, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group21, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group22, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group23, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group24, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group25, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group26, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group27, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group28, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(NA_factor ~ group29, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 

#Daily stressor: 
summary(lme(stress ~ group1, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group2, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group3, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group4, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group5, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group6, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group7, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 
summary(lme(stress ~ group8, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 1)) 

summary(lme(stress ~ group9, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(stress ~ group10, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(stress ~ group11, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(stress ~ group12, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(stress ~ group13, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 
summary(lme(stress ~ group14, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"))) 

summary(lme(stress ~ group15, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group16, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group17, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group18, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group19, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group20, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group21, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group22, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group23, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group24, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group25, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group26, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group27, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group28, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 
summary(lme(stress ~ group29, data = my_data, random=~1|idtw/subjno, na.action=na.omit, control=list(opt="optim"), subset = group1 != 2)) 

###And now multiverse analysis

test1 <- lme(NA_permean ~ l1_stress + id_time + group1 + l1_stress*group1 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group1 != 1)
test2 <- lme(NA_permean ~ l1_stress + id_time + group2 + l1_stress*group2 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group2 != 1)
test3 <- lme(NA_permean ~ l1_stress + id_time + group3 + l1_stress*group3 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group3 != 1)
test4 <- lme(NA_permean ~ l1_stress + id_time + group4 + l1_stress*group4 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group4 != 1)
test5 <- lme(NA_permean ~ l1_stress + id_time + group5 + l1_stress*group5 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group5 != 1)
test6 <- lme(NA_permean ~ l1_stress + id_time + group6 + l1_stress*group6 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group6 != 1)
test7 <- lme(NA_permean ~ l1_stress + id_time + group7 + l1_stress*group7 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group7 != 1)
test8 <- lme(NA_permean ~ l1_stress + id_time + group8 + l1_stress*group8 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group8 != 1)

test9 <- lme(NA_permean ~ l1_stress + id_time + group9 + l1_stress*group9 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
test10 <- lme(NA_permean ~ l1_stress + id_time + group10 + l1_stress*group10 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
test11 <- lme(NA_permean ~ l1_stress + id_time + group11 + l1_stress*group11 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
test12 <- lme(NA_permean ~ l1_stress + id_time + group12 + l1_stress*group12 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
test13 <- lme(NA_permean ~ l1_stress + id_time + group13 + l1_stress*group13 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))
test14 <- lme(NA_permean ~ l1_stress + id_time + group14 + l1_stress*group14 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno))

test15 <- lme(NA_permean ~ l1_stress + id_time + group15 + l1_stress*group15 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group15 != 2)
test16 <- lme(NA_permean ~ l1_stress + id_time + group16 + l1_stress*group16 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group16 != 2)
test17 <- lme(NA_permean ~ l1_stress + id_time + group17 + l1_stress*group17 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group17 != 2)
test18 <- lme(NA_permean ~ l1_stress + id_time + group18 + l1_stress*group18 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group18 != 2)
test19 <- lme(NA_permean ~ l1_stress + id_time + group19 + l1_stress*group19 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group19 != 2)
test20 <- lme(NA_permean ~ l1_stress + id_time + group20 + l1_stress*group20 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group20 != 2)
test21 <- lme(NA_permean ~ l1_stress + id_time + group21 + l1_stress*group21 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group21 != 2)
test22 <- lme(NA_permean ~ l1_stress + id_time + group22 + l1_stress*group22 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group22 != 2)
test23 <- lme(NA_permean ~ l1_stress + id_time + group23 + l1_stress*group23 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group23 != 2)
test24 <- lme(NA_permean ~ l1_stress + id_time + group24 + l1_stress*group24 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group24 != 2)
test25 <- lme(NA_permean ~ l1_stress + id_time + group25 + l1_stress*group25 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group25 != 2)
test26 <- lme(NA_permean ~ l1_stress + id_time + group26 + l1_stress*group26 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group26 != 2)
test27 <- lme(NA_permean ~ l1_stress + id_time + group27 + l1_stress*group27 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group27 != 2)
test28 <- lme(NA_permean ~ l1_stress + id_time + group28 + l1_stress*group28 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group28 != 2)
test29 <- lme(NA_permean ~ l1_stress + id_time + group28 + l1_stress*group29 + gender + age_wave1, random=list(idtw = ~1, subjno = pdDiag(~ 1 + id_time + l1_stress)), data=my_data_complete, na.action=na.omit, control=list(opt="optim"), correlation = corCAR1(form = ~ id_time|idtw/subjno), subset = group29 != 2)


vector_of_p <- c(summary(test1)$tTable[7, 5], summary(test2)$tTable[7, 5], summary(test3)$tTable[7, 5], summary(test4)$tTable[7, 5],
                 summary(test5)$tTable[7, 5], summary(test6)$tTable[7, 5], summary(test7)$tTable[7, 5], summary(test8)$tTable[7, 5],
                 summary(test9)$tTable[7, 5], summary(test10)$tTable[7, 5], summary(test11)$tTable[7, 5], summary(test12)$tTable[7, 5],
                 summary(test13)$tTable[7, 5], summary(test14)$tTable[7, 5], summary(test15)$tTable[7, 5], summary(test16)$tTable[7, 5],
                 summary(test17)$tTable[7, 5], summary(test18)$tTable[7, 5], summary(test19)$tTable[7, 5], summary(test20)$tTable[7, 5],
                 summary(test21)$tTable[7, 5], summary(test22)$tTable[7, 5], summary(test23)$tTable[7, 5], summary(test24)$tTable[7, 5],
                 summary(test25)$tTable[7, 5], summary(test26)$tTable[7, 5], summary(test27)$tTable[7, 5], summary(test28)$tTable[7, 5],
                 summary(test29)$tTable[7, 5])
hist(vector_of_p, breaks = 10, col = "grey", main = "Distribution of p-values", xlab = "p-value")
abline(v=0.05, col="red")