C&S Rep

Load some useful packages.

library(lme4) # mixed model regressions

## Loading required package: Matrix
## Loading required package: Rcpp

library(bootstrap) # bootstraps for confidence intervals
library(ggplot2) # plotting
library(plyr) # data manipulation

rm(list=ls())

## for bootstrapping 95% confidence intervals
theta <- function(x,xdata,na.rm=T) {mean(xdata[x],na.rm=na.rm)}
ci.low <- function(x,na.rm=T) {
  mean(x,na.rm=na.rm) - quantile(bootstrap(1:length(x),1000,theta,x,na.rm=na.rm)$thetastar,.025,na.rm=na.rm)}
ci.high <- function(x,na.rm=T) {
  quantile(bootstrap(1:length(x),1000,theta,x,na.rm=na.rm)$thetastar,.975,na.rm=na.rm) - mean(x,na.rm=na.rm)}

Read in data.

d <- read.csv("~/Desktop/C&S/results-8-22-14.csv")
 
# reorder factor levels for condition
d$Input.condition <- factor(d$Input.condition, 
                             levels=c("false","local","literal","all"))

First filter participants based on the two manipulation checks.

mean(d$Input.letters_check == as.numeric(as.character(d$Answer.letters_check)),
     na.rm=TRUE)

## [1] 0.9646

mean(d$Input.circles_check == as.numeric(as.character(d$Answer.circles_check)),
     na.rm=TRUE)

## [1] 0.89

dc <- subset(d, 
             d$Input.letters_check == 
               as.numeric(as.character(d$Answer.letters_check)) &
               d$Input.circles_check == 
               as.numeric(as.character(d$Answer.circles_check)))

## Warning: NAs introduced by coercion

Finally, check for repeats in the data.

ordered <- dc[order(dc$SubmitTime),]
dcc <- ordered[!duplicated(ordered$WorkerId),]

Now see how much we’ve lost.

length(d$WorkerId)

## [1] 200

length(dc$WorkerId)

## [1] 169

length(dcc$WorkerId)

## [1] 130

Main analysis

Aggregate data across conditions.

ms <- ddply(dcc, .(Input.condition), summarise,
      rating = mean(Answer.truth_judgment),
      cih = ci.high(Answer.truth_judgment),
      cil = ci.low(Answer.truth_judgment))

Now plot means.

theme_set(theme_bw()) # set ggplot2 theme to be a little nicer

qplot(Input.condition, rating, 
      ymin=rating-cil, ymax=rating+cih, 
      geom="pointrange", 
      data=ms) + 
  scale_y_continuous(breaks=c(1,3,5,7)) +
  ylab("Mean Rating") + 
  xlab("Condition")

Also look at histograms over individual responses.

qplot(Answer.truth_judgment, facets=.~Input.condition, binwidth=1,
      data=dcc) + 
  scale_x_continuous(breaks=c(1,3,5,7)) + 
  xlab("Truth Judgment")