Study1_Judgments_of_Policy_Efficacy_after_the_Learning_Task_Causal.R

#######################################
# Judgment Accuracy: Causal Functions #
#######################################

#import libraries
library('rcompanion') # 'wilcoxonOneSampleR()' 
library("tidyverse")

# Import data
Study1_Judgments_of_Policy_Efficacy_after_the_Learning_Task_Causal <- as_tibble(read.csv("Study1_Judgments_of_Policy_Efficacy_after_the_Learning_Task_Causal.csv"))

# Preview data
head(Study1_Judgments_of_Policy_Efficacy_after_the_Learning_Task_Causal)

# data dictionary:
## "Participant" = participant number
## "Function" = function number (causal functions only; range 1-4; refer to payoff function figure)
## "Congruence" = denotes preference-congruence for a given policy/function, 1=yes, -1=no (i.e., is preferred policy actually the best policy)
## "Ambiguity" = denotes ambiguity of function (low vs high)
## "Judgment_Accuracy" = values are the distance a final judgment was from the correct assessment value; lower scores reflect greater accuracy (range: 0-10). 

# Note: As explained in the paper, two separate analyses were conducted for each predictor variable (congruence and ambiguity). These analyses are divided below. 

#######################
# Congruence Analysis #
#######################

jpe_causal_congruence <- Study1_Judgments_of_Policy_Efficacy_after_the_Learning_Task_Causal %>% 
  group_by(Participant, Congruence) %>% 
  summarise(Mean_Judgment_Accuracy = mean(Judgment_Accuracy, na.rm=T))


jpe_causal_preference_congruent <- jpe_causal_congruence %>% 
  filter(Congruence==1)

jpe_causal_preference_incongruent <- jpe_causal_congruence %>% 
  filter(Congruence==-1)

# descriptives
jpe_causal_preference_congruent %>% 
  summary() # Mean_Judgment_Accuracy median = 3

jpe_causal_preference_incongruent %>% 
  summary() # Mean_Judgment_Accuracy median = 6


# Wilcoxon Signed Rank Tests
wilcox.test(x=jpe_causal_preference_congruent$Mean_Judgment_Accuracy,
            y=jpe_causal_preference_incongruent$Mean_Judgment_Accuracy,
            paired=FALSE,
            exact=TRUE
)
# W = 354.50, p < .001

# effect size: r
wilcoxonR(x=jpe_causal_congruence$Mean_Judgment_Accuracy,
          g=jpe_causal_congruence$Congruence,
          ci=TRUE, paired=TRUE) # r= .44, CI = .249 – .620 (Note: CI resamples every time it is run)


######################
# Ambiguity Analysis #
######################

jpe_causal_ambiguity <- Study1_Judgments_of_Policy_Efficacy_after_the_Learning_Task_Causal %>% 
  group_by(Participant, Ambiguity) %>% 
  summarise(Mean_Judgment_Accuracy = mean(Judgment_Accuracy, na.rm=T))


jpe_causal_ambiguity_low <- jpe_causal_ambiguity %>% 
  filter(Ambiguity=="Low")

jpe_causal_ambiguity_high <- jpe_causal_ambiguity %>% 
  filter(Ambiguity=="High")

# descriptives
jpe_causal_ambiguity_low %>% 
  summary() # Mean_Judgment_Accuracy median = 2

jpe_causal_ambiguity_high %>% 
  summary() # Mean_Judgment_Accuracy median = 5


# Wilcoxon Signed Rank Tests
wilcox.test(x=jpe_causal_ambiguity_low$Mean_Judgment_Accuracy,
            y=jpe_causal_ambiguity_high$Mean_Judgment_Accuracy,
            paired=TRUE,
            exact=TRUE
)
# W = 94.50, p < .001

# effect size: r
wilcoxonR(x=jpe_causal_ambiguity$Mean_Judgment_Accuracy,
          g=jpe_causal_ambiguity$Ambiguity,
          ci=TRUE, paired=TRUE) # r= .49, CI = .32 – .65 (Note: CI resamples every time it is run)