Prep

Reading in Libraries and Data
#### Libraries
library(stargazer)
library("psych")
library("tidyr")
library(lmerTest)
library(lme4)
library(optimx)
#install.packages("optimx")
library(sjstats)
library(psycho)
library(dplyr)
library(foreign)
library(nlme)
library(sjPlot)
#install.packages("sjPlot")
library(ggplot2)
library(ggeffects)
library("mlVAR")
library("graphicalVAR")
library("qgraph")
library("scales")
library("PerformanceAnalytics")
library("tidyverse")
library("ppcor")
library("igraph")
library("mediation")
library("lavaan")
library("gridExtra")
library("ggcorrplot")
require(plyr)
library(magrittr)
library(kableExtra)
library(viridis)
    #•  'Category' is a variable that codes whether the phrase is a 1 = Generalized anxiety phrase, 2 = depressive phrase, or 3 = social anxiety phrase (there are substantially fewer SAD phrases)
    #•  'Response.RESP' refers to whether the subject said the word and the sentence either 1 = matched or 3 = did not match
    #•  'Type' is a variable that codes whether the word displayed was 1 = non-negative or 2 = negative
wsap <- read.csv("/Users/nikki/Desktop/Research/WSAP/WSAP_uncleaned.csv")
wsap_l <- read.csv("/Users/nikki/Desktop/Research/WSAP/wsap_long_cleaned_outs_72221.csv")
wsap_l <- wsap_l[-c(1)]
wsap_w <- read.csv("/Users/nikki/Desktop/Research/WSAP/wsap_wide_cleaned_outs_72221.csv")
wsap_w <- wsap_w[-c(1)]


trait <- read.csv("/Users/nikki/Desktop/Research/HealthyU_Scanning_Local/Emily_HUresources/WRG Survey/HealthyU_Questionnaires_Masterdata_Wide.csv", header=T)
trait <- trait[c(2,13:18,25:30)]

#EMA and GPS variables
ema <- read.csv("/Users/nikki/Desktop/Research/EMA/HealthyU_EMA_andGPS_onlyEMAdays_AllData.csv")
ema <- ema[c(1,3,8,9,10:19,32,33,46,40,42)]

Data cleaning, org, calculating out comes


6 people were removed for having < 83% accuracy on practice trials (missed 2 or more out of 6)

5033 5036 5047 5049 5050 5081

167 Trials were removed for being > 3SD longer or shorter than the SUBJECTS mean

Then, 2 subjects were removed for being > SD longer, on average, than the average of all subjects

##Practice Trials
wsap$Response1.RESP <- factor(wsap$Response1.RESP,
                              labels = c("related","unrelated"),
                              levels = c(1,3))

prac_wsap <- subset(wsap, wsap$Block ==1)
prac_wsap$Category <- as.factor(as.character(prac_wsap$Category))

#calculate accuracy
prac_wsap$correct <- prac_wsap$Category == prac_wsap$Response1.RESP

#merge back into main data frame
acc <-aggregate(prac_wsap$correct,list(prac_wsap$Subject),mean,na.rm=TRUE)
names(acc)<-c("Subject","prac_acc") 
prac_wsap <-merge(prac_wsap,acc,by="Subject") 
##Task Trials

#merge practice accuracy scores for exclusion
wsap <- subset(wsap, wsap$Block ==2)

wsap <- merge(wsap,acc,by="Subject") 
wsap_lowacc <- subset(wsap, wsap$prac_acc < 0.83) #length(unique(wsap_lowacc$Subject)) = 6 people

#exclude subjects with more than 1/6 incorrect practice trials
wsap <- subset(wsap, wsap$prac_acc >= 0.83) #length(unique(wsap$Subject)) = 71 people

#factorize the responses
  wsap$response <- factor(wsap$Response.RESP,
                                labels = c("related","unrelated"),
                                levels = c(1,3))
  
  wsap$wordvalence <- factor(wsap$Type_1p_2n, 
                               labels = c("not negative","negative"),
                               levels = c(1,2))
  
  wsap$Category <- as.numeric(as.character(wsap$Category))
  wsap$trialtype <- factor(wsap$Category, 
                             labels = c("gad","dep","sad"),
                             levels = c(1,2,3))
#reduce dataframe
wsap_longer <- wsap
wsap <- wsap[c(1,2,4,13,18,21,22,20,19)]
  # Check and exclude RTs > 3SD: first WITHIN SUBs
    grprts <-aggregate(wsap$Response.RT,list(wsap$Subject),mean,na.rm=TRUE)
    grprtsds <-aggregate(wsap$Response.RT,list(wsap$Subject),sd,na.rm=TRUE)

    names(grprts)<-c("Subject","subrt") #rename variables
    names(grprtsds)<-c("Subject","subrtsds") #rename variables
    wsap<-merge(wsap,grprts,by="Subject") #merge group means into original data
    wsap<-merge(wsap,grprtsds,by="Subject")
    wsap$cent_RT<- (wsap$Response.RT-wsap$subrt) # group mean center
    wsap$norm_RT<- wsap$cent_RT/wsap$subrtsds # group mean center
    length(which(wsap$norm_RT >= 3 | wsap$norm_RT <= -3))
    exclude <- which(wsap$norm_RT > 3 | wsap$norm_RT < -3)

    wsap_allRTs <- wsap
    wsap <- wsap[-c(exclude),]
    
    grprts2 <-aggregate(wsap$Response.RT,list(wsap$Subject),mean,na.rm=TRUE)
    names(grprts2)<-c("Subject","subrt2") #rename variables
    wsap<-merge(wsap,grprts2,by="Subject") #merge group means into original data
    wsap$norm_subrt2 <- (wsap$subrt2- (mean(grprts2$subrt2)))/sd(grprts2$subrt2)
    #wsap$Subject[which(wsap$norm_subrt2 > 3)]
    
    wsap_noRTsubjectexl <- wsap
    wsap <- wsap[-c(which(wsap$norm_subrt2 > 3)),]
    #unique(wsap$Subject) #69 total
#calc outcome of interest
wsap$Threat_end <- ifelse(wsap$wordvalence == "negative" & wsap$response == "related", 1, 0)
wsap$Threat_rej <- ifelse(wsap$wordvalence == "negative" & wsap$response == "unrelated", 1, 0)
wsap$Benign_end <- ifelse(wsap$wordvalence == "not negative" & wsap$response == "related", 1, 0)
wsap$Benign_rej <- ifelse(wsap$wordvalence == "not negative" & wsap$response == "unrelated", 1, 0)

wsap$RT_Threat_end <- ifelse(wsap$wordvalence == "negative" & wsap$response == "related", wsap$Response.RT, NA)
wsap$RT_Threat_rej <- ifelse(wsap$wordvalence == "negative" & wsap$response == "unrelated", wsap$Response.RT, NA)
wsap$RT_Benign_end<- ifelse(wsap$wordvalence == "not negative" & wsap$response == "related", wsap$Response.RT, NA)
wsap$RT_Benign_rej <- ifelse(wsap$wordvalence == "not negative" & wsap$response == "unrelated", wsap$Response.RT, NA)
#subject choice averages and rates
subcount_TE <- as.data.frame(with(wsap, tapply(Threat_end, Subject, sum, na.rm = T)))
subcount_trials <- as.data.frame(with(wsap, tapply(Trial, Subject, length)))
colnames(subcount_TE) <- c("subcount_TE")
colnames(subcount_trials) <- c("subcount_trials")
subcount_TE$subcount_TE_rate <- subcount_TE$subcount_TE/subcount_trials$subcount_trials

subcount_TR <- as.data.frame(with(wsap, tapply(Threat_rej, Subject, sum, na.rm = T)))
colnames(subcount_TR) <- c("subcount_TR")
subcount_TR$subcount_TR_rate <- subcount_TR$subcount_TR/subcount_trials$subcount_trials

subcount_BE <- as.data.frame(with(wsap, tapply(Benign_end, Subject, sum, na.rm = T)))
colnames(subcount_BE) <- c("subcount_BE")
subcount_BE$subcount_BE_rate <- subcount_BE$subcount_BE/subcount_trials$subcount_trials

subcount_BR <- as.data.frame(with(wsap, tapply(Benign_rej, Subject, sum, na.rm = T)))
colnames(subcount_BR) <- c("subcount_BR")
subcount_BR$subcount_BR_rate <- subcount_BR$subcount_BR/subcount_trials$subcount_trials

wsap_rates <- cbind(subcount_TE,subcount_TR,subcount_BE,subcount_BR)
wsap_rates$Subjects <- row.names(wsap_rates)

#   d'?
wsap_rates$threat_d <- qnorm(wsap_rates$subcount_BE_rate) - qnorm(wsap_rates$subcount_TE_rate)
subRT_TE <- as.data.frame(with(wsap, tapply(RT_Threat_end, Subject, mean, na.rm = T)))
colnames(subRT_TE) <- c("subRT_TE")

subRT_TR <- as.data.frame(with(wsap, tapply(RT_Threat_rej, Subject, mean, na.rm = T)))
colnames(subRT_TR) <- c("subRT_TR")

subRT_BE <- as.data.frame(with(wsap, tapply(RT_Benign_end, Subject, mean, na.rm = T)))
colnames(subRT_BE) <- c("subRT_BE")

subRT_BR <- as.data.frame(with(wsap, tapply(RT_Benign_rej, Subject, mean, na.rm = T)))
colnames(subRT_BR) <- c("subRT_BR")

wsap_rates2 <- cbind(subRT_TE,subRT_TR,subRT_BE,subRT_BR)
wsap_rates2$Subjects <- row.names(wsap_rates2)

wsap_rates <- merge(wsap_rates, wsap_rates2, by = "Subjects")
#write.csv(wsap, "/Users/nikki/Desktop/Research/WSAP/wsap_long_cleaned_outs_72221.csv")
#write.csv(wsap_rates, "/Users/nikki/Desktop/Research/WSAP/wsap_wide_cleaned_outs_72221.csv")
wsap_spr1 <- spread(wsap_l, trialtype, Threat_end)
wsap_spr2 <- spread(wsap_l, trialtype, Benign_end)
wsap_spr3 <- spread(wsap_l, trialtype, RT_Threat_end)
wsap_spr4 <- spread(wsap_l, trialtype, RT_Benign_end)

wsap_spread <- cbind(wsap_spr1[c(22:24)],wsap_spr2[c(22:24)],wsap_spr3[c(22:24)],wsap_spr4[c(22:24)])
colnames(wsap_spread) <- c("dep_TE", "gad_TE", "sad_TE","dep_BE", "gad_BE", "sad_BE","dep_TE_RT", "gad_TE_RT", "sad_TE_RT","dep_BE_RT", "gad_BE_RT", "sad_BE_RT")
wsap_spread$Subject <- wsap_l$Subject
wsap_spread$trialtype <- wsap_l$trialtype

#subject choice averages and rates
subcount_dep_TE <- as.data.frame(with(wsap_spread, tapply(dep_TE, Subject, sum, na.rm = T)))
colnames(subcount_dep_TE) <- c("subcount_dep_TE")

subcount_gad_TE <- as.data.frame(with(wsap_spread, tapply(gad_TE, Subject, sum, na.rm = T)))
colnames(subcount_gad_TE) <- c("subcount_gad_TE")

subcount_sad_TE <- as.data.frame(with(wsap_spread, tapply(sad_TE, Subject, sum, na.rm = T)))
colnames(subcount_sad_TE) <- c("subcount_sad_TE")

    subcount_dep_BE <- as.data.frame(with(wsap_spread, tapply(dep_BE, Subject, sum, na.rm = T)))
    colnames(subcount_dep_BE) <- c("subcount_dep_BE")
    
    subcount_gad_BE <- as.data.frame(with(wsap_spread, tapply(gad_BE, Subject, sum, na.rm = T)))
    colnames(subcount_gad_BE) <- c("subcount_gad_BE")
    
    subcount_sad_BE <- as.data.frame(with(wsap_spread, tapply(sad_BE, Subject, sum, na.rm = T)))
    colnames(subcount_sad_BE) <- c("subcount_sad_BE")

subcount_dep_TE_RT <- as.data.frame(with(wsap_spread, tapply(dep_TE_RT, Subject, mean, na.rm = T)))
colnames(subcount_dep_TE_RT) <- c("subcount_dep_TE_RT")

subcount_gad_TE_RT <- as.data.frame(with(wsap_spread, tapply(gad_TE_RT, Subject, mean, na.rm = T)))
colnames(subcount_gad_TE_RT) <- c("subcount_gad_TE_RT")

subcount_sad_TE_RT <- as.data.frame(with(wsap_spread, tapply(sad_TE_RT, Subject, mean, na.rm = T)))
colnames(subcount_sad_TE_RT) <- c("subcount_sad_TE_RT")

    subcount_dep_BE_RT <- as.data.frame(with(wsap_spread, tapply(dep_BE_RT, Subject, mean, na.rm = T)))
    colnames(subcount_dep_BE_RT) <- c("subcount_dep_BE_RT")
    
    subcount_gad_BE_RT <- as.data.frame(with(wsap_spread, tapply(gad_BE_RT, Subject, mean, na.rm = T)))
    colnames(subcount_gad_BE_RT) <- c("subcount_gad_BE_RT")
    
    subcount_sad_BE_RT <- as.data.frame(with(wsap_spread, tapply(sad_BE_RT, Subject, mean, na.rm = T)))
    colnames(subcount_sad_BE_RT) <- c("subcount_sad_BE_RT")

wsap_specific <- cbind(subcount_dep_TE, subcount_gad_TE, subcount_sad_TE, subcount_dep_BE, subcount_gad_BE, subcount_sad_BE, subcount_dep_TE_RT, subcount_gad_TE_RT, subcount_sad_TE_RT, subcount_dep_BE_RT, subcount_gad_BE_RT, subcount_sad_BE_RT)
                               
wsap_specific$Subjects <- wsap_w$Subjects

#wsap_outcomes <- merge(wsap_w, wsap_specific, by = "Subjects")

#subcount_trials <- as.data.frame(with(wsap_l, tapply(Trial, Subject, length)))
#wsap_outcomes$numTrials <- subcount_trials$`with(wsap_l, tapply(Trial, Subject, length))`
#write.csv(wsap_outcomes, "/Users/nikki/Desktop/Research/WSAP/wsap_wide_cleaned_outs_72221.csv")

Reliability…tbd

Initial look at WSAP & Symptoms

Task Details

Objective

Subjects see a word for 500ms then they see a sentence and must decide whether the word is “related” or “unrelated” to the sentence



Trials

There are 140 trials: 80 depression, 40 general anxiety, 20 social anxiety*

Subjects see every sentence twice, once with a negative/threat word and once with a benign word



Common Outcomes

“Offline” = choice rates, or the number of times they endorse or reject divided by the possible # of times “Online” = choice reaction time

Thus, choices fall into 1 of 4 categories: Threat Endorse, Threat Reject, Benign Endorse, Benign Reject



Basic Plotting and Visualization

wsap_outcomes <- wsap_w
colors <- c("Threat Endorsed" = "lightpink2", "Threat Rejected" = "plum3", "Benign Endorsed" = "lightblue3", "Benign Rejected" = "wheat2")
ggplot(data = wsap_outcomes) +
       geom_density(aes(x = subcount_TE_rate, fill = "Threat Endorsed"), alpha = .3) +
       geom_density(aes(x = subcount_TR_rate, fill = "Threat Rejected"), alpha = .3) +
       geom_density(aes(x = subcount_BE_rate, fill = "Benign Endorsed"), alpha = .3) +
       geom_density(aes(x = subcount_BR_rate, fill = "Benign Rejected"), alpha = .3) +
       labs(title = "Density of Choice Rates across Sample",
         x = "Rate",
         y = "Density",
         color = "Legend") +
         scale_color_manual(values = colors) +
         scale_fill_discrete(name = "Type of Choice") + 
       theme_gray()

ggplot(data = wsap_outcomes) +
  geom_histogram(aes(x = threat_d), color = "black", fill = "gold3", binwidth = .10, alpha = .75) + 
  labs(title = "Distribution of d' for threat detection",
         x = "d' (larger + values reflect less sensitivity to threat)",
         y = "Count") +
  theme_minimal()

colors <- c("Threat Endorsed" = "lightpink2", "Threat Rejected" = "plum3", "Benign Endorsed" = "lightblue3", "Benign Rejected" = "wheat2")
ggplot(data = wsap_outcomes) +
       geom_density(aes(x = subRT_TE, fill = "Threat Endorsed"), alpha = .5) +
       geom_density(aes(x = subRT_TR, fill = "Threat Rejected"), alpha = .5) +
       geom_density(aes(x = subRT_BE, fill = "Benign Endorsed"), alpha = .5) +
       geom_density(aes(x = subRT_BR, fill = "Benign Rejected"), alpha = .5) +
       labs(title = "Density of Reaction Times for Specific Choices across Sample",
         x = "Reaction Time",
         y = "Density",
         color = "Legend") +
         scale_color_manual(values = colors) +
         scale_fill_discrete(name = "Type of Choice") + 
       theme_gray()

wsap_outcomes$subcount_dep_TE_sc <- as.vector(scale(wsap_outcomes$subcount_dep_TE))
wsap_outcomes$subcount_dep_BE_sc <- as.vector(scale(wsap_outcomes$subcount_dep_BE))
wsap_outcomes$subcount_gad_TE_sc <- as.vector(scale(wsap_outcomes$subcount_gad_TE))
wsap_outcomes$subcount_gad_BE_sc <- as.vector(scale(wsap_outcomes$subcount_gad_BE))
wsap_outcomes$subcount_sad_TE_sc <- as.vector(scale(wsap_outcomes$subcount_sad_TE))
wsap_outcomes$subcount_sad_BE_sc <- as.vector(scale(wsap_outcomes$subcount_sad_BE))
keycol <- "trialtype"
valuecol <- "rate"
gathercols <- c("subcount_dep_TE_sc","subcount_gad_TE_sc","subcount_sad_TE_sc","subcount_dep_BE_sc","subcount_gad_BE_sc","subcount_sad_BE_sc")
wsap_outcomes_long <- gather_(wsap_outcomes, keycol, valuecol, gathercols)
wsap_outcomes_long$trialtype <- as.factor(wsap_outcomes_long$trialtype)

ggplot(data = wsap_outcomes_long, aes(x = trialtype, y =  rate, fill = trialtype)) +
  geom_violin(alpha = .85) +
  labs(title = "Distribution of Normalized Endorsement Rates across Trial types",
         x = "Types of choices across Trial types",
         y = "") +
         scale_fill_manual(name="Endorsement x Trial type",
                           labels=c("Benign/Dep", "Threat/Dep", "Benign/Gad","Threat/Gad","Benign/Sad","Threat/Sad"),
                            values = c("coral3","lightsalmon1","palegreen4", "palegreen", "skyblue4","skyblue")) +
   theme(axis.text.x=element_blank(), axis.ticks.x=element_blank())

Descriptives

var_of_int <- wsap_w[c(27,10,3,5,7,9,15:20,11:14,21:26)]
var_sum <- describe(var_of_int)
var_sum  <- round(var_sum , digits=2)
#Knit a lil table
var_sum %>%
  kbl() %>%
  kable_styling()
vars n mean sd median trimmed mad min max range skew kurtosis se
numTrials 1 69 137.67 1.13 138.00 137.70 1.48 135.00 140.00 5.00 -0.23 -0.11 0.14
threat_d 2 69 0.37 0.27 0.36 0.38 0.23 -0.43 1.02 1.45 -0.33 0.69 0.03
subcount_TE_rate 3 69 0.24 0.08 0.23 0.24 0.08 0.09 0.44 0.34 0.43 -0.44 0.01
subcount_TR_rate 4 69 0.26 0.08 0.27 0.26 0.08 0.06 0.41 0.35 -0.40 -0.36 0.01
subcount_BE_rate 5 69 0.36 0.06 0.37 0.36 0.05 0.20 0.46 0.27 -0.64 0.06 0.01
subcount_BR_rate 6 69 0.14 0.06 0.13 0.14 0.05 0.04 0.30 0.26 0.66 -0.05 0.01
subcount_dep_TE 7 69 17.19 6.40 16.00 16.82 5.93 2.00 35.00 33.00 0.52 0.30 0.77
subcount_gad_TE 8 69 8.80 3.06 9.00 8.75 2.97 3.00 15.00 12.00 0.14 -0.86 0.37
subcount_sad_TE 9 69 6.97 2.64 7.00 6.95 2.97 2.00 12.00 10.00 0.01 -1.02 0.32
subcount_dep_BE 10 69 28.07 4.52 28.00 28.30 4.45 15.00 36.00 21.00 -0.39 -0.20 0.54
subcount_gad_BE 11 69 10.35 2.30 11.00 10.54 1.48 3.00 15.00 12.00 -0.87 1.04 0.28
subcount_sad_BE 12 69 11.07 2.61 12.00 11.28 1.48 4.00 15.00 11.00 -0.71 -0.47 0.31
subRT_TE 13 69 887.67 334.60 885.50 868.66 323.15 340.53 1739.53 1399.00 0.52 -0.37 40.28
subRT_TR 14 69 947.81 413.32 953.63 919.19 386.56 306.03 2584.54 2278.52 1.27 3.29 49.76
subRT_BE 15 69 829.35 321.64 793.26 808.99 289.35 308.45 1751.59 1443.13 0.58 0.00 38.72
subRT_BR 16 69 1060.50 464.95 984.50 1023.89 432.18 360.25 2351.47 1991.22 0.75 0.07 55.97
subcount_dep_TE_RT 17 69 913.35 379.47 880.92 887.37 395.98 321.47 1999.89 1678.42 0.66 -0.16 45.68
subcount_gad_TE_RT 18 69 831.77 354.67 774.25 797.70 338.90 334.79 2272.40 1937.61 1.25 2.54 42.70
subcount_sad_TE_RT 19 69 897.93 420.85 811.75 855.92 391.96 283.00 2264.50 1981.50 0.98 0.82 50.66
subcount_dep_BE_RT 20 69 843.74 331.15 848.71 819.15 318.72 303.35 1744.47 1441.13 0.70 0.23 39.87
subcount_gad_BE_RT 21 69 830.92 374.97 802.80 810.55 393.33 271.08 1758.82 1487.73 0.44 -0.72 45.14
subcount_sad_BE_RT 22 69 791.70 322.54 746.80 769.31 323.82 278.25 1872.43 1594.18 0.76 0.47 38.83

Basic Correlations

cordat1 <- wsap_outcomes[c(10,3,5,7,9)]
colnames(cordat1) <- c("d'","Overall Threat Endorse","Overall Threat Reject", "Overall Benign Endorse","Overall Benign Reject")
cormat1 <- correlation(cordat1)
r1 <- as.data.frame(cormat1$values$r)
p1 <- as.data.frame(cormat1$values$p)
p.mat1 <- cor_pmat(cordat1)

 ggcorrplot(r1, method = "square", ggtheme = ggplot2::theme_minimal, title = "Correlations between broad, general Choice Rates", show.legend = TRUE, legend.title = "Corr", show.diag = FALSE,
  colors = c("skyblue4","white", "palevioletred4"), outline.color = "black",
  hc.order = F, hc.method = "pairwise", lab = T,
  lab_col = "black", lab_size = 3, p.mat = p.mat1, sig.level = 0.05,
  insig = c("pch"), pch = 7, pch.col = "black",
  pch.cex = 16, tl.cex = 11, tl.col = "black", tl.srt = 45,
  digits = 2)

cordat2 <- wsap_outcomes[c(3,15:26)]
colnames(cordat2) <- c("Overall Threat Endorse","Dep: Threat Endorse","Gad: Threat Endorse","Sad: Threat Endorse","Dep: Benign Endorse", "Gad: Benign Endorse","Sad: Benign Endorse","RT for Dep Threat Endorse","RT for Gad Threat Endorse", "RT for Sad Threat Endorse","RT for Dep Benign Endorse","RT for Gad Benign Endorse", "RT for Sad Benign Endorse")
cormat2 <- correlation(cordat2)
r2 <- as.data.frame(cormat2$values$r)
p2 <- as.data.frame(cormat2$values$p)
p.mat2 <- cor_pmat(cordat2)

  ggcorrplot(r2, method = "square", ggtheme = ggplot2::theme_minimal, title = "Correlations between Dx-specific Choice Rates and RTs", show.legend = TRUE, legend.title = "Corr", show.diag = FALSE,
  colors = c("skyblue4","white", "palevioletred4"), outline.color = "black",
  hc.order = F, hc.method = "pairwise", lab = T,
  lab_col = "black", lab_size = 2, p.mat = p.mat2, sig.level = 0.05,
  insig = c("pch"), pch = 7, pch.col = "black",
  pch.cex = 4, tl.cex = 11, tl.col = "black", tl.srt = 45,
  digits = 2)


Basic Task Models of Choices and RT

Found 3 things:

1. main effect of valence with LESS endorsement of negative words compared to benign words
2. main effect of trial type with LESS endorsement of depressive words compared to general anx words
3. main effect of trial type with LESS endorsement of social anx words compared to general anx words




Also tested the effects of word valence (positive/negative) and trial type (dep/gad/sad trials) on a person’s reaction time for their trial choice.

###Basic Task MLMs CONT.

#Implicit/Online RT
RT_model1 <- lmer(Response.RT ~ wordvalence + trialtype + (1|Subject), data = wsap)
#summary(RT_model1)
tab_model(RT_model1)
  Response.RT
Predictors Estimates CI p
(Intercept) 916.87 836.36 – 997.39 <0.001
wordvalence [not
negative] 		-8.66 -34.57 – 17.24 0.512
trialtype [gad] -39.16 -71.97 – -6.34 0.019
trialtype [sad] -65.72 -98.47 – -32.97 <0.001
Random Effects
σ2 414773.13
τ00 Subject 108136.13
ICC 0.21
N Subject 69
Observations 9499
Marginal R2 / Conditional R2 0.001 / 0.208 
  #only main effect is depression trials show SLOWER choices

#Adding in more random effects
RT_model2 <- (lmer(Response.RT ~ wordvalence + trialtype + (1 + wordvalence|Subject), data = wsap))
#summary(model2)
 
 #not sig improvement
 #anova(RT_model1, RT_model2)


Symptom Measure Distributions

healthyu <- merge(trait, wsap_w, by.x = "id", by.y = "Subjects")
healthyu$threat_d <- qnorm(healthyu$subcount_TE_rate) - qnorm(healthyu$subcount_BE_rate)

healthyu$GADtot_T3_sc <- scale(healthyu$GADtot_T3)
healthyu$PHQ9_tot_T3_sc <- scale(healthyu$PHQ9_tot_T3)
healthyu$SIAS_tot_T3_sc <- scale(healthyu$SIAS_tot_T3)
healthyu$OCIR_tot_T3_sc <- scale(healthyu$OCIR_tot_T3)
healthyu$SIR_tot_T3_sc <- scale(healthyu$SIR_tot_T3)
healthyu$EPSI_tot_T3_sc <- scale(healthyu$EPSI_tot_T3)

a <- ggplot(data = healthyu, aes(x = PHQ9_tot_T3_sc)) +
  #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
  geom_density(fill = "palevioletred2", alpha = .35) +
             labs(title = "Density of Depression Symptoms",
         x = "Depression Symptoms (PHQ9)",
         y = "Density") +
       theme_gray() + 
       theme(plot.title = element_text(size = 8),
             axis.title.x = element_text(size = 6))

b <- ggplot(data = healthyu, aes(x = GADtot_T3_sc)) +
  #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
  geom_density(fill = "darkorange2", alpha = .35) +
             labs(title = "Density of GAD Symptoms",
         x = "General Anxiety Symptoms (GAD7)",
         y = "Density") +
       theme_gray() + 
       theme(plot.title = element_text(size = 8),
             axis.title.x = element_text(size = 6))

c <- ggplot(data = healthyu, aes(x = SIAS_tot_T3_sc)) +
  #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
  geom_density(fill = "darkgoldenrod2", alpha = .35) +
             labs(title = "Density of SAD Symptoms",
         x = "Social Anxiety Symptoms (SIAS)",
         y = "Density") +
       theme_gray() + 
       theme(plot.title = element_text(size = 8),
             axis.title.x = element_text(size = 6))

d <- ggplot(data = healthyu, aes(x = OCIR_tot_T3_sc)) +
  #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
  geom_density(fill = "palegreen2", alpha = .35) +
             labs(title = "Density of OC Symptoms",
         x = "Obsessive Compulsive Symptoms (OCIR)",
         y = "Density") +
       theme_gray() + 
       theme(plot.title = element_text(size = 8),
             axis.title.x = element_text(size = 6))

e <- ggplot(data = healthyu, aes(x = SIR_tot_T3_sc)) +
  #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
  geom_density(fill = "deepskyblue2", alpha = .35) +
             labs(title = "Density of Hoarding Symptoms",
         x = "Hoarding Symptoms (SIR)",
         y = "Density") +
       theme_gray() + 
       theme(plot.title = element_text(size = 8),
             axis.title.x = element_text(size = 6))

f <- ggplot(data = healthyu, aes(x = EPSI_tot_T3_sc)) +
  #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
  geom_density(fill = "mediumpurple3", alpha = .35) +
             labs(title = "Density of Eating Disorder Symptoms",
         x = "Eating Disorder Symptoms (EPSI)",
         y = "Density") +
       theme_gray() + 
       theme(plot.title = element_text(size = 8),
             axis.title.x = element_text(size = 6))

grid.arrange(a,b,c,d,e,f, nrow = 2)

# a <- ggplot(data = healthyu, aes(x = PHQ9_tot_T3_sc)) +
#   #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
#   geom_density(fill = "dodgerblue2", alpha = .35) +
#              labs(title = "Density of Despression Symptoms",
#          x = "Depression Symptoms (PHQ9)",
#          y = "Density") +
#        theme_gray() + 
#        theme(plot.title = element_text(size = 8),
#              axis.title.x = element_text(size = 6))
# 
# b <- ggplot(data = healthyu, aes(x = GADtot_T3_sc)) +
#   #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
#   geom_density(fill = "skyblue3", alpha = .35) +
#              labs(title = "Density of GAD Symptoms",
#          x = "General Anxiety Symptoms (GAD7)",
#          y = "Density") +
#        theme_gray() + 
#        theme(plot.title = element_text(size = 8),
#              axis.title.x = element_text(size = 6))
# 
# c <- ggplot(data = healthyu, aes(x = SIAS_tot_T3_sc)) +
#   #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
#   geom_density(fill = "lightblue1", alpha = .35) +
#              labs(title = "Density of SAD Symptoms",
#          x = "Social Anxiety Symptoms (SIAS)",
#          y = "Density") +
#        theme_gray() + 
#        theme(plot.title = element_text(size = 8),
#              axis.title.x = element_text(size = 6))
# 
# d <- ggplot(data = healthyu, aes(x = OCIR_tot_T3_sc)) +
#   #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
#   geom_density(fill = "mediumseagreen", alpha = .35) +
#              labs(title = "Density of OC Symptoms",
#          x = "Obsessive Compulsive Symptoms (OCIR)",
#          y = "Density") +
#        theme_gray() + 
#        theme(plot.title = element_text(size = 8),
#              axis.title.x = element_text(size = 6))
# 
# e <- ggplot(data = healthyu, aes(x = SIR_tot_T3_sc)) +
#   #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
#   geom_density(fill = "palegreen3", alpha = .35) +
#              labs(title = "Density of Hoarding Symptoms",
#          x = "Hoarding Symptoms (SIR)",
#          y = "Density") +
#        theme_gray() + 
#        theme(plot.title = element_text(size = 8),
#              axis.title.x = element_text(size = 6))
# 
# f <- ggplot(data = healthyu, aes(x = EPSI_tot_T3_sc)) +
#   #geom_histogram(aes(x = PHQ9_tot_T3_sc), binwidth = .25) +
#   geom_density(fill = "darkolivegreen3", alpha = .35) +
#              labs(title = "Density of Eating Disorder Symptoms",
#          x = "Eating Disorder Symptoms (EPSI)",
#          y = "Density") +
#        theme_gray() + 
#        theme(plot.title = element_text(size = 8),
#              axis.title.x = element_text(size = 6))
# 
# grid.arrange(a,b,c,d,e,f, nrow = 2)

Analysis of Main Aims:



Aim 1


Does a broad metric of interp. bias from the WSAP relate to symptoms transdiagnostically?


Brief primer to signal detection and d’
d’ was calculated to reflect one’s threat sensitivity/bias
Here, a Hit is endorsing a negative/threat word and a False alarm is endorsing a benign word



healthyu <- merge(trait, wsap_w, by.x = "id", by.y = "Subjects")
healthyu$threat_d <- qnorm(healthyu$subcount_TE_rate) - qnorm(healthyu$subcount_BE_rate)

healthyu$GADtot_T3_sc <- scale(healthyu$GADtot_T3)
healthyu$PHQ9_tot_T3_sc <- scale(healthyu$PHQ9_tot_T3)
healthyu$SIAS_tot_T3_sc <- scale(healthyu$SIAS_tot_T3)
healthyu$OCIR_tot_T3_sc <- scale(healthyu$OCIR_tot_T3)
healthyu$SIR_tot_T3_sc <- scale(healthyu$SIR_tot_T3)
healthyu$EPSI_tot_T3_sc <- scale(healthyu$EPSI_tot_T3)

gen.wsap.sx <- healthyu[c(40:45,22,15,23,19,25)]
# gen.wsap.sx.fdr <- gen.wsap.sx %>% 
#   correlation(adjust = "fdr", i_am_cheating = T)
# summary(gen.wsap.sx.fdr)

colnames(gen.wsap.sx) <- c("GAD sx","Dep sx","SAD sx","OC sx","Hoarding sx","Eating sx","d'","Overall Threat Endorse","RT for Threat Endorse", "Overall Benign Endorse","RT for Benign Endorse")
cormat1 <- correlation(gen.wsap.sx)
r1 <- as.data.frame(cormat1$values$r)
p1 <- as.data.frame(cormat1$values$p)
p.mat1 <- cor_pmat(gen.wsap.sx)

 ggcorrplot(r1, method = "square", ggtheme = ggplot2::theme_minimal, title = "Correlations b/w Symptoms & General WSAP bias", show.legend = TRUE, legend.title = "Corr", show.diag = FALSE,
  colors = c("skyblue4","white", "palevioletred4"), outline.color = "black",
  hc.order = F, hc.method = "pairwise", lab = T,
  lab_col = "black", lab_size = 2, p.mat = p.mat1, sig.level = 0.05,
  insig = c("pch"), pch = 7, pch.col = "black",
  pch.cex = 8, tl.cex = 10, tl.col = "black", tl.srt = 45,
  digits = 2)

Overall Threat Sensitivity/Bias is linked to all Symptom measures

#"dodgerblue2","skyblue3","lightblue1","mediumseagreen","palegreen3","darkolivegreen3"
#TE <- ggplot(data = healthyu) +
#   geom_point(aes(x = threat_d, y = GADtot_T3_sc), color = "skyblue3", alpha=0.35,) + 
#   geom_point(aes(x = threat_d, y = PHQ9_tot_T3_sc), color = "dodgerblue2",alpha=0.35,) + 
#   geom_point(aes(x = threat_d, y = OCIR_tot_T3_sc), color = "mediumseagreen",alpha=0.35,) + 
#   geom_point(aes(x = threat_d, y = SIR_tot_T3_sc), color = "palegreen3",alpha=0.35,) + 
#   geom_point(aes(x = threat_d, y = EPSI_tot_T3_sc), color = "darkolivegreen3",alpha=0.35,) + 
#   geom_point(aes(x = threat_d, y = SIAS_tot_T3_sc), color = "lightblue1",alpha=0.35,) + 
#     geom_smooth(aes(x = threat_d, y = GADtot_T3_sc), color = "skyblue3", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = threat_d, y = PHQ9_tot_T3_sc), color = "dodgerblue2", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = threat_d, y = OCIR_tot_T3_sc), color = "mediumseagreen", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = threat_d, y = SIR_tot_T3_sc), color = "palegreen3", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = threat_d, y = EPSI_tot_T3_sc), color = "darkolivegreen3", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = threat_d, y = SIAS_tot_T3_sc), color = "lightblue1", size = 1, alpha=0.25, method = "lm", se = F) +
#     labs(title= "Significant associations between threat endorsement and all symptom measures", x = "Sensitivity to Threat Words", y = "Symptom Measures")+
#   theme_minimal()
# 
# TE

TE <- ggplot(data = healthyu) +
  geom_point(aes(x = threat_d, y = GADtot_T3_sc), color = "darkorange2", alpha=0.35,) + 
  geom_point(aes(x = threat_d, y = PHQ9_tot_T3_sc), color = "palevioletred2",alpha=0.35,) + 
  geom_point(aes(x = threat_d, y = OCIR_tot_T3_sc), color = "palegreen2",alpha=0.35,) + 
  geom_point(aes(x = threat_d, y = SIR_tot_T3_sc), color = "deepskyblue2",alpha=0.35,) + 
  geom_point(aes(x = threat_d, y = EPSI_tot_T3_sc), color = "mediumpurple3",alpha=0.35,) + 
  geom_point(aes(x = threat_d, y = SIAS_tot_T3_sc), color = "darkgoldenrod2",alpha=0.35,) + 
    geom_smooth(aes(x = threat_d, y = GADtot_T3_sc), color = "darkorange2", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = threat_d, y = PHQ9_tot_T3_sc), color = "palevioletred2", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = threat_d, y = OCIR_tot_T3_sc), color = "palegreen2", size = 1, alpha=0.25, method = "lm", se = F) +
   geom_smooth(aes(x = threat_d, y = SIR_tot_T3_sc), color = "deepskyblue2", size = 1, alpha=0.25, method = "lm", se = F) +
   geom_smooth(aes(x = threat_d, y = EPSI_tot_T3_sc), color = "mediumpurple3", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = threat_d, y = SIAS_tot_T3_sc), color = "darkgoldenrod2", size = 1, alpha=0.25, method = "lm", se = F) +
    labs(title= "Significant associations between threat endorsement and all symptom measures", x = "Sensitivity to Threat Words", y = "Symptom Measures")+
  theme_minimal()

TE

#cor.test(healthyu$GADtot_T3,healthyu$threat_d)
#cor.test(healthyu$PHQ9_tot_T3,healthyu$threat_d)
#cor.test(healthyu$OCIR_tot_T3,healthyu$threat_d)
#cor.test(healthyu$SIAS_T3,healthyu$threat_d)
#cor.test(healthyu$EPSI_tot_T3,healthyu$threat_d)
#cor.test(healthyu$SIAS_tot_T3,healthyu$threat_d)
TE_RT <- ggplot(data = healthyu) +
  geom_point(aes(x = subRT_TE, y = GADtot_T3_sc), color = "darkorange2", alpha=0.35,) + 
  geom_point(aes(x = subRT_TE, y = PHQ9_tot_T3_sc), color = "palevioletred2",alpha=0.35,) + 
  geom_point(aes(x = subRT_TE, y = OCIR_tot_T3_sc), color = "palegreen2",alpha=0.35,) + 
  geom_point(aes(x = subRT_TE, y = SIR_tot_T3_sc), color = "deepskyblue2",alpha=0.35,) + 
  geom_point(aes(x = subRT_TE, y = EPSI_tot_T3_sc), color = "mediumpurple3",alpha=0.35,) + 
  geom_point(aes(x = subRT_TE, y = SIAS_tot_T3_sc), color = "darkgoldenrod2",alpha=0.35,) + 
    geom_smooth(aes(x = subRT_TE, y = GADtot_T3_sc), color = "darkorange2", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = subRT_TE, y = PHQ9_tot_T3_sc), color = "palevioletred2", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = subRT_TE, y = OCIR_tot_T3_sc), color = "palegreen2", size = 2, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = subRT_TE, y = SIR_tot_T3_sc), color = "deepskyblue2", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = subRT_TE, y = EPSI_tot_T3_sc), color = "mediumpurple3", size = 1, alpha=0.25, method = "lm", se = F) +
  geom_smooth(aes(x = subRT_TE, y = SIAS_tot_T3_sc), color = "darkgoldenrod2", size = 1, alpha=0.25, method = "lm", se = F) +
  labs(title= "No significant associations b/w RT for threat endorsement and symptoms",
       x = "RT for Threat Words Endorsed", y = "Symptom Measures")+
  theme_minimal()

TE_RT

# task <- healthyu[c(1,22,15,19,23,25)]
# sx <- healthyu[c(1,40:45)]
# keycol <- "sx_type"
# valuecol <- "sx_score"
# gathercols <- c("GADtot_T3_sc","PHQ9_tot_T3_sc","SIAS_tot_T3_sc","OCIR_tot_T3_sc",
#                 "SIR_tot_T3_sc","EPSI_tot_T3_sc")
# sx_l <- gather_(sx, keycol, valuecol, gathercols)
# 
# wsap_sx <- merge(sx_l,task,by = "id")
# 
# model1 <- (lmer(sx_score ~ sx_type + subcount_TE_rate + (1 | id), data = wsap_sx))
# summary(model1)
# 
# model4 <- (lmer(threat_d ~ wordvalence + trialtype + (1 + wordvalence|Subject), data = wsap))



Aim 2


Do metrics of disorder-specific interp. bias from the WSAP relate to the specific symptoms they target?



Specific Bias markers (choice behavior) show some differential relationships to Symptom measures

- SAD, GAD, and DEP (not OC,HD,or ED) are ~ LOWER rate of benign endorsement on social anx. trials
- Only GAD and DEP ~ HIGHER rate of threat endorsement on general anx. trials
- All Sx ~ HIGHER rate of threat endorsement on depressive and social anx. trials
spec.wsap.sx <- healthyu[c(40:45,27:32)]
# gen.wsap.sx.fdr <- gen.wsap.sx %>% 
#   correlation(adjust = "fdr", i_am_cheating = T)
# summary(gen.wsap.sx.fdr)

colnames(spec.wsap.sx) <- c("GAD sx","Dep sx","SAD sx","OC sx","Hoarding sx","Eating sx","Depression: Threat Endorse", "General Anx: Threat Endorse","Social Anx: Threat Endorse",  "Depression: Benign Endorse", "General Anx: Benign Endorse","Social Anx: Benign Endorse")
cormat1 <- correlation(spec.wsap.sx)
r1 <- as.data.frame(cormat1$values$r)
p1 <- as.data.frame(cormat1$values$p)
p.mat1 <- cor_pmat(spec.wsap.sx)

 ggcorrplot(r1, method = "square", ggtheme = ggplot2::theme_minimal, title = "Correlations b/w Symptoms & Specific WSAP bias (choices)", show.legend = TRUE, legend.title = "Corr", show.diag = FALSE,
  colors = c("skyblue4","white", "palevioletred4"), outline.color = "black",
  hc.order = F, hc.method = "pairwise", lab = T,
  lab_col = "black", lab_size = 2, p.mat = p.mat1, sig.level = 0.05,
  insig = c("pch"), pch = 7, pch.col = "black",
  pch.cex = 6, tl.cex = 10, tl.col = "black", tl.srt = 45,
  digits = 2)

healthyu$subcount_dep_TE_resc <- rescale(healthyu$subcount_dep_TE, mean = 0, sd = 1,df=TRUE)
healthyu$subcount_gad_TE_resc <- rescale(healthyu$subcount_gad_TE, mean = 0, sd = 1,df=TRUE)
healthyu$subcount_sad_TE_resc <- rescale(healthyu$subcount_sad_TE, mean = 0, sd = 1,df=TRUE)

# 
#    ggplot(data = healthyu) +
#   geom_smooth(aes(x = subcount_dep_TE_resc, y = GADtot_T3_sc), color = "palevioletred1", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_dep_TE_resc, y = PHQ9_tot_T3_sc), color = "darkorange1", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_dep_TE_resc, y = OCIR_tot_T3_sc), color = "palegreen1", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = subcount_dep_TE_resc, y = SIR_tot_T3_sc), color = "deepskyblue1", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = subcount_dep_TE_resc, y = EPSI_tot_T3_sc), color = "mediumpurple1", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_dep_TE_resc, y = SIAS_tot_T3_sc), color = "darkgoldenrod1", size = 1, alpha=0.25, method = "lm", se = F) +
# # ylim(c(-3, 3)) + 
# #    labs(title= "Significant associations between threat endorsement on Depression trials and all symptom measures", x = "Sensitivity to Threat Words on Depression Trials", y = "Symptom Measures")+
# #  theme_minimal()
# 
# #GadSpecific_TE <- ggplot(data = healthyu) +
#   geom_smooth(aes(x = subcount_gad_TE_resc, y = GADtot_T3_sc), color = "palevioletred2", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_gad_TE_resc, y = PHQ9_tot_T3_sc), color = "darkorange2", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_gad_TE_resc, y = OCIR_tot_T3_sc), color = "palegreen2", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = subcount_gad_TE_resc, y = SIR_tot_T3_sc), color = "deepskyblue2", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = subcount_gad_TE_resc, y = EPSI_tot_T3_sc), color = "mediumpurple2", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_gad_TE_resc, y = SIAS_tot_T3_sc), color = "darkgoldenrod2", size = 1, alpha=0.25, method = "lm", se = F) +
# #  ylim(c(-3, 3)) + 
# #    labs(title= "Associations between threat endorsement on General Anx. trials and all symptom measures", x = "Sensitivity to Threat Words on General Anx. Trials", y = "Symptom Measures")+
# #  theme_minimal()
# 
# #SadSpecific_TE <- ggplot(data = healthyu) +
#   geom_smooth(aes(x = subcount_sad_TE_resc, y = GADtot_T3_sc), color = "palevioletred3", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_sad_TE_resc, y = PHQ9_tot_T3_sc), color = "darkorange3", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_sad_TE_resc, y = OCIR_tot_T3_sc), color = "palegreen3", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = subcount_sad_TE_resc, y = SIR_tot_T3_sc), color = "deepskyblue3", size = 1, alpha=0.25, method = "lm", se = F) +
#    geom_smooth(aes(x = subcount_sad_TE_resc, y = EPSI_tot_T3_sc), color = "mediumpurple3", size = 1, alpha=0.25, method = "lm", se = F) +
#   geom_smooth(aes(x = subcount_sad_TE_resc, y = SIAS_tot_T3_sc), color = "darkgoldenrod3", size = 1, alpha=0.25, method = "lm", se = F) +
# 
#   theme_minimal()
sxs <- healthyu[c(1,46:48)]
keycol <- "trial_type"
valuecol <- "endorse"
gathercols <- c("subcount_dep_TE_resc","subcount_gad_TE_resc","subcount_sad_TE_resc")
sxs_l <- gather_(sxs, keycol, valuecol, gathercols)
sxs_l_task <- merge(sxs_l, healthyu, by = "id")


keycol <- "sx_type"
valuecol <- "sx_score"
gathercols <- c("GADtot_T3_sc","PHQ9_tot_T3_sc","SIAS_tot_T3_sc")
sxs_l_task_l <- gather_(sxs_l_task, keycol, valuecol, gathercols)

sxs_l_task_l <- sxs_l_task_l[c(1,2,3,48:49)]
colnames(sxs_l_task_l) <- c("id","Trial_Type","endorse","Symptom_Measures","sx_score")
sxs_l_task_l$Trial_Type <- factor(sxs_l_task_l$Trial_Type,
                                  levels = c("subcount_dep_TE_resc","subcount_sad_TE_resc","subcount_gad_TE_resc"),
                                  labels = c("DEP Trials","SAD Trials","GAD Trials"))

sxs_l_task_l$Symptom_Measures <- factor(sxs_l_task_l$Symptom_Measures,
                                  levels = c("PHQ9_tot_T3_sc","SIAS_tot_T3_sc","GADtot_T3_sc"),
                                  labels = c("DEP Symptoms","SAD Symptoms", "GAD Symptoms"))


  ggplot(data = sxs_l_task_l, aes(x = endorse, y = sx_score, color = Symptom_Measures)) +
  geom_smooth(size = 1.5, method = "lm", se = F) +
    labs(title= "Associations between Dx-specific threat endorsement and symptom measures", subtitle = "     Only nonsig. effect is SAD sx and GAD threat endrosement", x = "Dx-specific threat endorsement", y = "Symptom Measures") +
    scale_color_viridis(discrete = TRUE, begin = .20, end = .85, option = "D")+
     facet_wrap(~ Trial_Type) +
    theme(legend.position="top") + 
    scale_fill_discrete(name = "Symptom Measures", labels = c("GAD", "DEP", "SAD"))



Aim 3


Does the broad, transdiagnostic interp. bias metricfrom the WSAP in the lab reflect patterns of affect in daily life?

# prep EMA data
    grpmeans<-aggregate(ema$NA_Mean_new,list(ema$subid),mean,na.rm=TRUE)
    names(grpmeans)<-c("Subject","ema_avg_na") #rename variables

    grpmeans_pa<-aggregate(ema$PA_Mean_tri,list(ema$subid),mean,na.rm=TRUE)
    names(grpmeans_pa)<-c("Subject","ema_avg_pa") #rename variables
    ema_means <-merge(grpmeans,grpmeans_pa,by="Subject") #merge group means into original data

    grpmeans_str<-aggregate(ema$allstress,list(ema$subid),mean,na.rm=TRUE)
    names(grpmeans_str)<-c("Subject","ema_all_str") #rename variables
    ema_means <-merge(ema_means,grpmeans_str,by="Subject") #merge group means into original data
    grpmeans_ent<-aggregate(ema$entropy_sameday,list(ema$subid),mean,na.rm=TRUE)
    names(grpmeans_ent)<-c("Subject","ema_entropy") #rename variables
    ema_means <-merge(ema_means,grpmeans_ent,by="Subject") #merge group means into original data
    grpmeans_soc<-aggregate(ema$SocConnect ,list(ema$subid),mean,na.rm=TRUE)
    names(grpmeans_soc)<-c("Subject","ema_soc") #rename variables
    ema_means <-merge(ema_means,grpmeans_soc,by="Subject") #merge group means into original data

    ## now suppose we want to extract the coefficients by group
    func <- function(ema)
    {
    return(data.frame(nastress = cor(ema$NA_Mean_new, ema$allstress, use = "pairwise.complete.obs")))
    }

    nastress2 <- ddply(ema, .(subid), func)
    ema_means <- merge(ema_means,nastress2,by.x = "Subject", by.y = "subid", all.x = T)

    
    ema_wsap <- merge(healthyu, ema_means, by.x = "id", by.y = "Subject", all = T)
ema_wsap$ema_all_str_resc <- scales::rescale(ema_wsap$ema_all_str, to = c(0,100), from = c(6,42))


  ggplot(data = ema_wsap) +
  geom_density(aes(x = ema_all_str_resc),fill = "darkolivegreen3", alpha = .5) +
  geom_density(aes(x = ema_avg_na),fill = "mediumpurple3", alpha = .5) +
  geom_density(aes(x = ema_avg_pa),fill = "gold3", alpha = .5) +
             labs(title = "Density of EMA Metrics",
         x = "Stress (green)           Negative Affect (purple)          Positive Affect (gold)",
         y = "Density") +
       theme_classic()

#Does the WSAP bias predict daily affect and stress in daily life? 
cordata <- ema_wsap[,c(22,15,49:51,54)]
cormat1 <- correlation(cordata, adjust = "fdr")
p.mat1 <- cor_pmat(cordata)

r1 <- as.data.frame(cormat1$values$r)
p1 <- as.data.frame(cormat1$values$p)
p.mat1 <- cor_pmat(cordata)

 ggcorrplot(r1, method = "square", ggtheme = ggplot2::theme_minimal, title = "Correlations b/w EMA & Broad WSAP bias", show.legend = TRUE, legend.title = "Corr", show.diag = FALSE,
  colors = c("skyblue4","white", "palevioletred4"), outline.color = "black",
  hc.order = F, hc.method = "pairwise", lab = T, 
  lab_col = "black", lab_size = 2, p.mat = p.mat1, sig.level = 0.05,
  insig = c("pch"), pch = 7, pch.col = "black",
  pch.cex = 14, tl.cex = 10, tl.col = "black", tl.srt = 45,
  digits = 2)

#stargazer(correlation.matrix,type = "html", p.auto = T, title="Correlation Matrix")

Specific Threat tnterpretation bias markers show some differential relationships daily stress and affect