Skip to content

Latest commit

 

History

History
175 lines (125 loc) · 5.74 KB

analysis.md

File metadata and controls

175 lines (125 loc) · 5.74 KB
Raw

Analysis for Epistemic Overconfidence in Algorithmic News Selection Paper

Scripts

Required Packages & Reproducibility

rm(list=ls())

renv::snapshot()
## * The lockfile is already up to date.

source("../lib/functions.R")

Load Data

load("../../data/intermediate/cleaned_data.RData") 
df <- df %>%
  mutate(esc = round(esc, 0),
         ent = round(ent, 0),
         hs = round(hs, 0),
         pt = round(pt, 0),
         surv = round(surv, 0))
#  pid = factor(pid, levels = c("None", "Other", "Democrat",
#                                      "Republican")))

Analysis

H1: Direct Effect of UGT on Algorithmic Appreciation

#df <- within(df, pid <- relevel(pid, ref = "Other"))
H1 <- lm(algo_app ~ esc + ent + hs + pt + surv +
           eo + factor(missing_eo) + news + polef + trust +
           age + factor(gender) + factor(country),  data = df)

H2: Interaction UGT * Epistemic Overconfidence

H2_1 <-  lm(algo_app ~ esc *eo + ent + hs + pt + surv +
            missing_eo + news + polef + trust + age +
            factor(gender) + factor(country),  data = df)

H2_2 <-  lm(algo_app ~ esc + ent*eo + hs + pt + surv +
              factor(missing_eo) + news + polef + trust + age +
              factor(gender) + factor(country),  data = df)

H2_3 <-  lm(algo_app ~ esc + ent + hs*eo + pt + surv +
              factor(missing_eo) + news + polef + trust +
              age + factor(gender) + factor(country),  data = df)

H2_4 <-  lm(algo_app ~ esc + ent + hs + pt*eo + surv +
              factor(missing_eo) + news + polef + trust + 
              age + factor(gender) + factor(country),  data = df)

H2_5 <-  lm(algo_app ~ esc + ent + hs + pt + surv*eo +
              factor(missing_eo) + news + polef + trust + age +
              factor(gender) + factor(country),  data = df)

Predicted Effects

Exploratory Analyses

Is Relationship between UGT and Epistemic Overconfidence Conditional upon Gender

Relationship between Appreciation for News Selector and Trust in Media

Ht1 <- lm(own_select ~ esc + ent + hs + pt + surv +
           eo + factor(missing_eo) + news + polef + trust +
           age + factor(gender) + factor(country),  data = df)

Ht2 <- lm(jou_select ~ esc + ent + hs + pt + surv +
           eo + factor(missing_eo) + news + polef + trust +
           age + factor(gender) + factor(country),  data = df)

Ht3 <- lm(trust ~ own_select + jou_select + eo +
          factor(missing_eo) + news + polef + age + 
          factor(gender) + factor(country),  data = df)

Relationship between Appreciation for Journalists as News Selectors and Overconfidence

Hj1 <-  lm(jou_select ~ esc *eo + ent + hs + pt + surv +
            missing_eo + news + polef + trust + age +
            factor(gender) + factor(country),  data = df)

Hj2 <-  lm(jou_select ~ esc + ent*eo + hs + pt + surv +
              factor(missing_eo) + news + polef + trust + age +
              factor(gender) + factor(country),  data = df)

Hj3 <-  lm(jou_select ~ esc + ent + hs*eo + pt + surv +
              factor(missing_eo) + news + polef + trust +
              age + factor(gender) + factor(country),  data = df)

Hj4 <-  lm(jou_select ~ esc + ent + hs + pt*eo + surv +
              factor(missing_eo) + news + polef + trust + 
              age + factor(gender) + factor(country),  data = df)

Hj5 <-  lm(jou_select ~ esc + ent + hs + pt + surv*eo +
              factor(missing_eo) + news + polef + trust + age +
              factor(gender) + factor(country),  data = df)

Predicted Effects

Relationship between Appreciation for Self Selecting the News and Overconfidence

Hs1 <-  lm(own_select ~ esc *eo + ent + hs + pt + surv +
            missing_eo + news + polef + trust + age +
            factor(gender) + factor(country),  data = df)

Hs2 <-  lm(own_select ~ esc + ent*eo + hs + pt + surv +
              factor(missing_eo) + news + polef + trust + age +
              factor(gender) + factor(country),  data = df)

Hs3 <-  lm(own_select ~ esc + ent + hs*eo + pt + surv +
              factor(missing_eo) + news + polef + trust +
              age + factor(gender) + factor(country),  data = df)

Hs4 <-  lm(own_select ~ esc + ent + hs + pt*eo + surv +
              factor(missing_eo) + news + polef + trust + 
              age + factor(gender) + factor(country),  data = df)

Hs5 <-  lm(own_select ~ esc + ent + hs + pt + surv*eo +
              factor(missing_eo) + news + polef + trust + age +
              factor(gender) + factor(country),  data = df)

Predicted Effects