7158_DissociationR.Rmd

title: "R Notebook"
output: html_notebook
---

# Notebook associated with analysis for CHI '22 paper 
# "I Don't Even Remember What I Read": How Design Influences
# Dissociation on Social Media
# By Amanda Baughan, Mingrui "Ray" Zhang, Raveena Rao, 
# Kai Lukoff, Anastasia Schaadhardt, Lisa Butler, and Alexis Hiniker
# https://doi.org/10.1145/3491102.3501899

## Read in dependencies, data, correct data types, etc. 
```{r echo=FALSE}
#If any of these do not load, try install.packages("lme4") (etc.), then try again
library(lme4)
library(lmerTest)
library(car)
library(emmeans)
library(multcomp)
library(coin)
library(zoo)
library(scales)
library(effectsize)

calculateBetas <- function(model){
    Vcov <- vcov(model, useScale = False)
    betas <- round(fixef(model), 3)#lmer
    se <- round(sqrt(diag(Vcov)), 3)
    zval <- round(betas / se, 3)
    pval <- round(2*pnorm(abs(zval), lower.tail = FALSE), 3)
    ## print everything
    cbind(betas, se, zval, pval)
}
```

```{r}
ldf <- read.csv("logs.csv")
#Set correct data types
ldf$pid <- as.factor(ldf$pid)
ldf$date <- as.factor(ldf$date)

length(unique(ldf$pid)) # number of participants
length(unique(ldf[ldf$avgAttentionScore > 3,]$pid)) # number of participants who experienced normative dissociation
length(ldf[ldf$avgAttentionScore > 3,]$avgAttentionScore) # instances of normative dissociation
```

# Internal interventions
## Read data and correct data types
```{r}
vtdf <- read.csv("views.csv")
vtdf$pid <- as.factor(vtdf$pid) #Participant ID
vtdf$date <- as.factor(vtdf$date) #Date recorded
vtdf$feedname_general <- as.factor(vtdf$feedname_general) #List or home feed
vtdf$internal <- as.factor(vtdf$internal) #Internal interventions on/off
vtdf$scrolledIntoOld <- as.factor(vtdf$scrolledIntoOld) #Scrolled into history yes/no
```

# Run an inverse gaussian model in which the "avgAttentionScore" (Dissociation score)
# is modeled by whether the feed is a list or home feed and the number of hours spent on it
# with the participant ID and the date as random variables
```{r}
vt.ig <- glmer(formula = avgAttentionScore ~ feedname_general + fthrs + (1|pid) + (1|date), family = inverse.gaussian(link = "log"), data =vtdf)
summary(vt.ig)
# Anova(vt.ig, type = "3")
```

# Run an inverse gaussian model in which dissociaiton is modeled as 
# the interaction of whether the reading history label would be shown at all (internal interventions on/off)
# and whether the participant scrolled into their old tweets, signaling that the
# reading history label has been seen
# PID and date are random variables
```{r}
rh.ig <- glmer(formula = avgAttentionScore ~ internal*scrolledIntoOld + (1|pid) + (1|date), family = inverse.gaussian(link = "log"), data =vtdf)
summary(rh.ig)
# Anova(rh.ig, type = "3")
```

```{r}
summary(glht(rh.ig, emm(pairwise ~ internal*scrolledIntoOld)), test=adjusted(type="holm"))
# 0, no reading history shown; 1, reading history is shown
# no, they do not scroll into previously read tweets; yes, they do scroll into previously read tweets
```

# External Interventions
# Model dissociation as part of total hours on Chirp, whether they used the time limit dialog to exit Chirp,
# whether they ignored the time limit dialog (dialogIgnore), and how many times they viewed usage stats
# pid and date are random variables
```{r}
m.ig <- glmer(formula = avgAttentionScore ~ consumetimeHours + dialogExit+dialogIgnore + statPageView + (1|pid) + (1|date), family = inverse.gaussian(link = "log"), data = ldf)
summary(m.ig)
# calculateBetas(m.ig)
# Anova(m.ig, type = "3")
```