BT 466 Final Code

setwd("C:/Users/debom/Documents/_Stevens/S2020/BT 466/R")
# Library imports
library(car)
library(data.table)
library(stringr)
library(multilevel)
library(mediation)

# Read input data
data <- read.csv("CSM Posts.csv", stringsAsFactors=FALSE, header=TRUE)

# Utiliity functions
replace_all <- function(df, pattern, replacement) {
  char <- vapply(df, function(x) is.factor(x) || is.character(x), logical(1))
  df[char] <- lapply(df[char], str_replace_all, pattern, replacement)
  df
}

# Data cleaning / preparation
data <- replace_all(data, ",", ".")
username <- data$Username
wc <- data$WC
data <- data[,12:ncol(data)]
data <- as.data.frame(sapply(data, as.numeric))
data <- cbind(username, data)
data <- cbind(wc, data)

# Adding custom data points for use in regression analysis
# Number of posts made (by user)
num_posts <- as.numeric(ave(username, username, FUN = length))
data <- cbind(num_posts, data)
cleaned <- data.frame(
  affiliation=scale(data$affiliation),
  username=data$username,
  num_posts=scale(data$num_posts),
  family=scale(data$family),
  friend=scale(data$friend),
  wc=scale(data$wc),
  social=scale(data$social),
  drives=scale(data$drives),
  achieve=scale(data$achieve),
  power=scale(data$power),
  percept=scale(data$percept),
  reward=scale(data$reward),
  home=scale(data$home)
)
# Moderator
social_drives <- cleaned$social * cleaned$drives
moderator <-lm(affiliation ~ social + drives + social_drives, data=cleaned)
summary(moderator)
# Indirect effect is 0.184
# Social is moderated by drives, low p-value -> significant

# Mediator
# Sobel arguments: (pred, med, out)
mediator <- sobel(data$family, data$social, data$affiliation)
mediator
p_value_2_sided <- 2*pnorm(-abs(mediator$z.value))
p_value_2_sided
# Indirect Effect: 0.219
# 2-Sided P-Value: 0 (significant?)
# Family is mediated by social
# Achievement is mediated by reward

# Multiple Regression
model <- lm(affiliation ~ . - username, data=cleaned)
options(scipen=999)
summary(model)
# R^2: 58.92%


# Moderated-Mediation
# This doesn't work at all

# Moderator: drives
low.drives <- mean(data$drives)-sd(data$drives)
high.drives <- mean(data$drives)+sd(data$drives)

# Mediator: social
Mod.Med.Model.1 <- lm(affiliation ~ drives * family, data=data)
Mod.Med.Model.2 <- lm(affiliation ~ drives * family + social, data=data)

Mod.Med.LowDrives <- mediate(Mod.Med.Model.1, Mod.Med.Model.2, covariates=list(drives=low.drives), boot = TRUE, boot.ci.type = "bca", sims = 10, treat="affiliation", mediator="social")
summary(Mod.Med.LowDrives)

Mod.Med.HighDrives <- mediate(Mod.Med.Model.1, Mod.Med.Model.2, covariates=list(drives=high.drives), boot = TRUE, boot.ci.type = "bca", sims = 10, treat="affiliation", mediator="social")
summary(Mod.Med.HighDrives)