-
Notifications
You must be signed in to change notification settings - Fork 0
/
exp3_analysis_main.Rmd
361 lines (290 loc) · 9.49 KB
/
exp3_analysis_main.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
---
title: 'Experiment 3: Main Analyses'
date: "`r Sys.Date()`"
output:
github_document:
toc: yes
toc_depth: 3
pdf_document:
toc: yes
toc_depth: 3
editor_options:
markdown:
wrap: 72
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(tidyverse)
options(dplyr.summarise.inform = FALSE)
library(magrittr)
library(lme4)
library(lmerTest)
library(broom.mixed)
library(insight)
library(kableExtra)
```
# Setup
Variable names:
- Experiment: exp3\_
- Data (\_d\_)
- d = main df
- count = sums of response types
- noOther = just *he* and *she* responses
- Models (\_m\_)
- all = effect of Condition and Name Gender Rating, including
*other* responses
- cond = effect of Condition only
- noOther = effect of Conditions (Last vs First+Full) and Name
Gender Rating, only on *he* and *she* responses
- FF = dummy coded with First + Full Name conditions as 0, Last
Name condition as 1
- L = dummy coded with Last Name condition as 0, First + Full Name
conditions as 1
Load data and select columns used in model. See data/exp3_data_about.txt
for more details.
```{r load-data}
exp3_d <- read.csv("../data/exp3_data.csv",
stringsAsFactors = TRUE) %>%
rename("Participant" = "SubjID", "Item" = "Name") %>%
select(
Participant, Condition, GenderRating,
Item, He, She, Other
)
str(exp3_d)
```
Center gender rating for names: Original scale from 1 to 7, with 1 as
most masculine and 7 as most feminine. Mean-centered with higher still
as more feminine.
```{r center-gender-rating}
exp3_d %<>% mutate(GenderRatingCentered = scale(GenderRating, scale = FALSE))
```
Set contrasts for name conditions, now weighted to account for uneven
sample sizes. This uses Scott Fraundorf's function for weighted
contrasts. (The psycholing package version doesn't support doing 2v1
comparisons, only 1v1.) Condition1 is Last vs First+Full. Condition2 is
First vs Full.
```{r contrast-coding}
source("centerfactor.R")
contrasts(exp3_d$Condition) <- centerfactor(
exp3_d$Condition, c("last", "first")
)
contrasts(exp3_d$Condition)
```
# Data Summary
Responses by condition.
```{r count-responses}
exp3_d %<>% mutate(ResponseAll = case_when(
He == 1 ~ "He",
She == 1 ~ "She",
Other == 1 ~ "Other"
))
exp3_d_count <- exp3_d %>%
group_by(Condition, ResponseAll) %>%
summarise(n = n()) %>%
pivot_wider(
names_from = ResponseAll,
values_from = n
) %>%
mutate(
She_HeOther = She / (He + Other),
She_He = She / He
) %>%
select(She, He, Other, She_HeOther, She_He)
kable(exp3_d_count, digits = 3)
```
# Model 1: With *Other* Responses
Effects of Condition (first name, last name, full name) and Gender
Rating on the likelihood of a *she* response, as opposed to a *he* or
*other* response. Participant and Item are included as random
intercepts, with items defined as the unique first, last and first +
last name combinations. Because the condition manipulations were fully
between-subject and between-item, fitting a random slope model was not
possible.
Because Experiment 3 always introduces the character with a full name,
then manipulates the name form in the subsequent 3 references, the main
analysis is one model, as opposed to the 2 for Experiment 1.
Condition1 is the contrast between last and first+full. Condition2 is
the contrast between first and full.
```{r model-all}
exp3_m_all <- glmer(
She ~ Condition * GenderRatingCentered + (1 | Participant) + (1 | Item),
data = exp3_d, family = binomial
)
summary(exp3_m_all)
```
- Fewer *she* responses overall
- Last Name vs First+Full Names condition effect only trending
- More *she* responses as first names become more feminine
- Larger effect of first name gender in First+Full Name conditions
than in Last Name conditions, which makes sense because there are 4
repetitions of the gendered first name, as opposed to only 1.
## Odds Ratios: Intercept
```{r OR-intercept-all}
exp(get_intercept(exp3_m_all))
exp(-get_intercept(exp3_m_all))
```
0.22x less likely to use *she* overall (or: 4.59x more likely to use
*he* and *other* overall), p\<.001
## Odds Ratios: Last vs First+Full
```{r OR-L-FF-all}
exp3_m_all %>%
tidy() %>%
filter(term == "Condition1") %>%
pull(estimate) %>%
exp()
```
1.17x more likely to use *she* than *he* and *other* in First + Full
compared to Last, p=0.09
## Odds Ratios: Last Only
Dummy code with Last Name as 0, so that intercept is the Last Name
condition only.
```{r dummy-code-L-all}
exp3_d %<>% mutate(Condition_Last = case_when(
Condition == "first" ~ 1,
Condition == "full" ~ 1,
Condition == "last" ~ 0
))
exp3_d$Condition_Last %<>% as.factor()
```
Model with just Condition (to more directly compare to Exp 1).
```{r model-L-all}
exp3_m_cond_L <- glmer(
She ~ Condition_Last + (1 | Participant) + (1 | Item),
data = exp3_d, family = binomial
)
summary(exp3_m_cond_L)
```
```{r OR-L-all}
exp(get_intercept(exp3_m_cond_L))
exp(-get_intercept(exp3_m_cond_L))
```
0.17x times less likely to use *she* than *he* and *other* in the Last
Name condition (or: 5.72x more likely to use *he* and *other* in the
Last Name condition), p\<.001
## Odds Ratios: First and Full Only
Dummy code with First and Full Name as 0, so the intercept is the
combination of those two.
```{r dummy-code-FF-all}
exp3_d %<>% mutate(Condition_FF = case_when(
Condition == "first" ~ 0,
Condition == "full" ~ 0,
Condition == "last" ~ 1
))
exp3_d$Condition_FF %<>% as.factor()
```
Model with just Condition (to more directly compare to Exp 1).
```{r model-FF-all}
exp3_m_cond_FF <- glmer(
She ~ Condition_FF + (1 | Participant) + (1 | Item),
data = exp3_d, family = binomial
)
summary(exp3_m_cond_FF)
```
```{r OR-FF-all}
exp(get_intercept(exp3_m_cond_FF))
exp(-get_intercept(exp3_m_cond_FF))
```
0.22x times less likely to use *she* than *he* and *other* in the First
and Full Name conditions (or: 4.46x more likely to use *he* and *other*
in the First and Full Name conditions), p\<.001
# Model 2: Without *Other* Responses
The sentence completion prompt for Experiment 3 is more open-ended than
in Experiment 1. So, we get a much higher proportion of *other*
responses (31% vs 7%), which I didn't anticipate.
```{r count-other}
sum(exp3_d$Other)
sum(exp3_d$Other) / length(exp3_d$Other)
```
```{r ubset-other}
exp3_d_noOther <- exp3_d %>% filter(Other == 0)
```
So, rerun the main model predicting the likelihood of *she* responses vs
*he* responses, with *other* responses excluded.
```{r model-other}
exp3_m_noOther <- glmer(
She ~ Condition * GenderRatingCentered + (1 | Participant) + (1 | Item),
data = exp3_d_noOther, family = binomial
)
summary(exp3_m_noOther)
```
These results are more similar to what we predicted from the previous
experiments:
- Fewer *she* responses overall
- Fewer *she* responses in the Last Name condition as compared to the
First + Full Name conditions (although we wouldn't predict as large
as a difference as in Exp1, because here there is one instance of
the first name in the Last Name condition)
- More *she* responses as first names become more feminine
- Larger effect of first name gender in First+Full Name conditions
than in Last Name conditions (which makes sense because there are
4repetitions of the gendered first name, as opposed to only 1.)
But, to keep the analyses consistent between experiments and avoid
post-hoc decision weirdness, both versions are reported.
## Odds Ratios: Intercept
```{r OR-intercept-other}
exp(get_intercept(exp3_m_noOther))
exp(-get_intercept(exp3_m_noOther))
```
0.65x less likely to use *she* than *he* overall (or: 1.53x more likely
to use *he* than *she* overall), p\<.001
## Odds Ratios: Last vs First+Full
```{r OR-L-FF-other}
exp3_m_noOther %>%
tidy() %>%
filter(term == "Condition1") %>%
pull(estimate) %>%
exp()
```
1.29x more likely to use *she* than *he* in First+Full than in Last (or:
1.29x more likely to use *he* than *she* in Last than in First+Full),
p\<.001
## Odds Ratios: Last Only
Dummy code with Last Name as 0, so that intercept is the Last Name
condition only.
```{r dummy-code-L-other}
exp3_d_noOther %<>% mutate(Condition_Last = case_when(
Condition == "first" ~ 1,
Condition == "full" ~ 1,
Condition == "last" ~ 0
))
exp3_d_noOther$Condition_Last %<>% as.factor()
```
```{r model-L-other}
exp3_m_noOther_L <- glmer(
She ~ Condition_Last + (1 | Participant) + (1 | Item),
data = exp3_d_noOther, family = binomial
)
summary(exp3_m_noOther_L)
```
```{r OR-L-other}
exp(get_intercept(exp3_m_noOther_L))
exp(-get_intercept(exp3_m_noOther_L))
```
0.51x times less likely to use *she* than *he* in the Last Name
condition (or: 1.97x more likely to use *he* than *she* in the Last Name
condition), p=.10
## Odds Ratios: First and Full Only
Dummy code with First and Full Name as 0, so the intercept is the
combination of those two.
```{r dummy-code-FF-other}
exp3_d_noOther %<>% mutate(Condition_FF = case_when(
Condition == "first" ~ 0,
Condition == "full" ~ 0,
Condition == "last" ~ 1
))
exp3_d_noOther$Condition_FF %<>% as.factor()
```
```{r model-FF-other}
exp3_m_noOther_FF <- glmer(
She ~ Condition_FF + (1 | Participant) + (1 | Item),
data = exp3_d_noOther, family = binomial
)
summary(exp3_m_noOther_FF)
```
```{r OR-FF-other}
exp(get_intercept(exp3_m_noOther_FF))
exp(-get_intercept(exp3_m_noOther_FF))
```
0.74x times less likely to use *she* than *he* and *other* in the First
and Full Name conditions (or: 1.35x more likely to use *he* and *other*
in the First and Full Name conditions), p=.46