exp4_analysis_supplementary.Rmd

---
title: 'Experiment 4: Supplementary Analyses'
date: "`r Sys.Date()`"
output: 
  github_document:
    toc: true
    toc_depth: 3
  pdf_document:
    toc: true
    toc_depth: 3
editor_options: 
  markdown: 
    wrap: 72
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(tidyverse)
options(dplyr.summarise.inform = FALSE)
library(magrittr)
library(lme4)
library(lmerTest)
library(buildmer)
library(insight)
library(broom.mixed)
library(kableExtra)
```

# Setup

Variable names:

-   Experiment: exp4\_

-   Data (\_d\_)

    -   d = main df
    -   noOther = just *male* and *female* responses

-   Models (\_m\_)

    -   noOther = effect of Conditions (Last vs First+Full) and Name
        Gender Rating, only on *male* and *female* responses
    -   FF = dummy coded with First + Full Name conditions as 0, Last
        Name condition as 1
    -   L = dummy coded with Last Name condition as 0, First + Full Name
        conditions as 1
    -   quad = quadratic effect of Name Gender
    -   subjGender = participant gender
    -   recenter= center name gender rating by scale (at 4)

Load data and select columns used in model. See data/exp4_data_about.txt
for more details.

```{r load-data}
exp4_d <- read.csv("../data/exp4_data.csv",
                   stringsAsFactors = TRUE) %>%
  rename("Participant" = "SubjID", "Item" = "Name") %>%
  select(
    Participant, Condition, SubjGenderMale,
    GenderRating, Item, Male, Female, Other
  )
str(exp4_d)
```

Center gender rating for names: Original scale from 1 to 7, with 1 as
most masculine and 7 as most feminine. Mean-centered with higher still
as more feminine.

```{r center-gender-rating}
exp4_d %<>% mutate(GenderRatingCentered = scale(GenderRating, scale = FALSE))
```

Set contrasts for name conditions, now weighted to account for uneven
sample sizes. This uses Scott Fraundorf's function for weighted
contrasts. (The psycholing package version doesn't support doing 2v1
comparisons, only 1v1.) Condition1 is Last vs First+Full. Condition2 is
First vs Full.

```{r contrast-coding}
source("centerfactor.R")
contrasts(exp4_d$Condition) <- centerfactor(exp4_d$Condition,
                                            c("last", "first"))
contrasts(exp4_d$Condition)
```

# Without *Other* Responses

The first supplementary analysis tests if excluding *other* responses
(2.99% of total responses) affects the pattern of results.

```{r count-other}
sum(exp4_d$Other)
sum(exp4_d$Other) / length(exp4_d$Other)
```

Exclude *other* responses.

```{r subset-other}
exp4_d_noOther <- exp4_d %>% filter(Other == 0)
```

Effect of Name Condition (first name, last name, full name) and first
name Gender Rating on likelihood of a *female* response, as opposed to a
*male* response, with *other* responses excluded. Participant and Item
are again included as random intercepts, with items defined as the
unique first, last and first + last name combinations.

```{r model-other}
exp4_m_noOther <- glmer(
  Female ~ Condition * GenderRatingCentered + (1 | Participant) + (1 | Item),
  data = exp4_d_noOther, family = binomial
)
summary(exp4_m_noOther)
```

Compared to the main model:

-   Intercept and Condition2:GenderRatingCentered (difference between
    Last Name and First+Full name conditions) potentially smaller
    differences

-   Condition2 now trending

## Odds Ratios: Intercept

```{r OR-intercept-other}
exp(get_intercept(exp4_m_noOther))
exp(-get_intercept(exp4_m_noOther))
```

0.84x less likely to recall as female overall (or: 1.18x more likely to
recall as male overall), p\<.05

## Odds Ratios: Last vs First+Full

```{r OR-L-FF-other}
exp4_m_noOther %>%
  tidy() %>%
  filter(term == "Condition1") %>%
  pull(estimate) %>%
  exp()
```

1.14x more likely to recall as female in First + Full compared to Last,
p\<.05

## Odds Ratios: Last Only

Dummy code with Last Name as 0, so that intercept is the Last Name
condition only.

```{r dummy-code-L-other}
exp4_d_noOther %<>% mutate(Condition_Last = case_when(
  Condition == "first" ~ 1,
  Condition == "full"  ~ 1,
  Condition == "last"  ~ 0
))
exp4_d_noOther$Condition_Last %<>% as.factor()
```

```{r model-L-other}
exp4_m_noOther_L <- glmer(
  Female ~ Condition_Last + (1 | Participant) + (1 | Item),
  data = exp4_d_noOther, family = binomial
)
summary(exp4_m_noOther_L)
```

```{r OR-L-other}
exp(get_intercept(exp4_m_noOther_L))
exp(-get_intercept(exp4_m_noOther_L))
```

0.76x times less likely to recall as female in the Last Name condition
(or: 1.31x more likely to recall as male in the Last Name condition),
p=.17

## Odds Ratios: First and Full Only

Dummy code with First and Full Name as 0, so that intercept is average
for these two conditions.

```{r dummy-code-FF-other}
exp4_d_noOther %<>% mutate(Condition_FF = case_when(
  Condition == "first" ~ 0,
  Condition == "full"  ~ 0,
  Condition == "last"  ~ 1
))
exp4_d_noOther$Condition_FF %<>% as.factor()
```

```{r model-FF-other}
exp4_m_noOther_FF <- glmer(
  Female ~ Condition_FF + (1 | Participant) + (1 | Item),
  data = exp4_d_noOther, family = binomial
)
summary(exp4_m_noOther_FF)
```

```{r OR-FF-other}
exp(get_intercept(exp4_m_noOther_FF))
exp(-get_intercept(exp4_m_noOther_FF))
```

0.89x less likely to recall as female in First and Full Name conditions
(or: 1.12x more likely to recall as male in First and Full Name
conditions), p=0.56

# Quadratic Name Gender Rating

The second supplementary analysis tested the effect of squared name
gender rating, such that larger values meant names with stronger gender
associations (masc or fem), and smaller values meant names with weaker
gender associations.

```{r model-quad}
exp4_d %<>% mutate(GenderRatingSquared = GenderRatingCentered^2)

exp4_m_quad <- glmer(
  Female ~ Condition * GenderRatingCentered + Condition * GenderRatingSquared +
    (1 | Participant) + (1 | Item),
  data = exp4_d, family = binomial
)
summary(exp4_m_quad)
```

-   Condition (F v F) \* Quadratic Gender Rating interaction, but n.s.
    after correction for multiple comparisons, so not making a big deal
    of it

# Participant Gender

## Setup/Data Summary

The third supplementary analysis looks at participant gender: if male
participants show a larger bias to recall the character as male than
non-male participants.

Participants entered their gender in a free-response box.

```{r count-subj-gender}
exp4_d %>%
  group_by(SubjGenderMale) %>%
  summarise(total = n_distinct(Participant)) %>%
  kable()
```

For this analysis, we exclude participants who did not respond (N=91).
Because there are not enough participants to create 3 groups, we compare
male to non-male participants. Male (N=602) and transgender male (N=1)
are coded as 1, and female (N=555), nonbinary (N=3), and transgender
female (N=1) are coded as 0.

Summary of responses by condition and participant gender:

```{r means-subj-gender}
exp4_d_subjGender <- exp4_d %>%
  filter(!is.na(SubjGenderMale)) %>%
  mutate(ResponseAll = case_when(
    Male   == 1 ~ "Male",
    Female == 1 ~ "Female",
    Other  == 1 ~ "Other"
  ))

exp4_d_subjGender %>%
  group_by(SubjGenderMale) %>%
  summarise(total = n_distinct(Participant)) %>%
  kable()
```

Participant gender is mean centered effects coded, comparing non-male
participants to male participants.

```{r contrasts-subj-gender}
exp4_d_subjGender$SubjGenderMale %<>% as.factor()
contrasts(exp4_d_subjGender$SubjGenderMale) <- cbind("NM_M" = c(-.5, .5))
contrasts(exp4_d_subjGender$SubjGenderMale)
```

## Model: Condition \* Name Gender \* Participant Gender

Effects of Name Condition (first name, full name), the first name's
Gender Rating (centered, positive=more feminine), and Participant Gender
(non-male vs. male) on the likelihood of a *female* response as opposed
to *male* or *other* responses.

```{r model-subj-gender}
exp4_m_subjGender <- glmer(
  Female ~ Condition * GenderRatingCentered * SubjGenderMale +
    (1 | Participant) + (1 | Item),
  data = exp4_d_subjGender, family = binomial
)
summary(exp4_m_subjGender)
```

-   Male participants less likely to recall character as female than
    non-male participants overall.

-   No other interactions with participant gender significant.

# Gender Rating Centering

The first name gender ratings aren't perfectly centered, partially
because mostly-feminine/somewhat-masculine names are much less common
than mostly-masculine/somewhat-feminine names.

```{r gender-rating-mean-center}
mean(exp4_d$GenderRating, na.rm = TRUE)
```

Does it make a difference if we center it on 4, the mean of the scale,
instead of 4.21, the mean of the items?

```{r gender-rating-abs-center}
exp4_d %<>% mutate(GenderRating4 = GenderRating - 4)
```

```{r model-gender-rating-recenter}
exp4_m_recenter <- glmer(
  Female ~ Condition * GenderRating4 + (1 | Participant) + (1 | Item),
  data = exp4_d, family = binomial
)
summary(exp4_m_recenter)
```

Here, the beta estimate for the intercept has a larger absolute value
(-0.41 vs -0.26), and the beta estimates for the condition effects is
slightly different (0.10 vs 0.13; 0.09 vs 0.07).