forked from tracymont/hyena_mobbing
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path16.model_participation_juveniles.R
320 lines (272 loc) · 14 KB
/
16.model_participation_juveniles.R
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
######################################################################
##### 16.0 Modeling mobbing participation - juveniles #####
######################################################################
########## 16.1 Set working directory & download packages/tables ##########
rm(list = ls())
setwd("~/Documents/R/LionHyena")
options(stringsAsFactors = FALSE)
library(GGally)
library(glmmTMB)
library(MuMIn)
library(multcomp)
library(DHARMa)
library(see)
library(performance)
library(broom.mixed)
library(viridis)
library(sjPlot)
library(patchwork)
library(tidyverse)
hyenadata::update_tables("1.2.88")
library(hyenadata)
#Load data
load("12.id_by_mob.Rdata")
#Set plot colors using viridis
viridis_2 <- viridis(7)[-c(1,3,5,6,7)]
viridis_3 <- viridis(7)[-c(2,4,6,7)]
#Function glht : https://cran.r-project.org/web/packages/glmmTMB/vignettes/model_evaluation.html
glht_glmmTMB <- function (model, ..., component="cond") {
glht(model, ...,
coef. = function(x) fixef(x)[[component]],
vcov. = function(x) vcov(x)[[component]],
df = NULL)
}
#Create dataset
id.by.mob.juv <- filter(id.by.mob, age.class == "juvenile") #1170
########## 16.2 Looking at combined data ##########
#Session-level
table(id.by.mob.juv$mobber)
table(id.by.mob.juv$context)
#Outliers
dotchart(id.by.mob.juv$age)
dotchart(id.by.mob.juv$stan.rank)
dotchart(id.by.mob.juv$prop.mob.related)
dotchart(id.by.mob.juv$mob.ai.avg)
dotchart(id.by.mob.juv$present.prop.rank.higher)
#Distribution/zeros
hist(id.by.mob.juv$age)
hist(id.by.mob.juv$stan.rank)
hist(id.by.mob.juv$prop.mob.related)
hist(id.by.mob.juv$mob.ai.avg)
hist(id.by.mob.juv$present.prop.rank.higher)
table(id.by.mob.juv$grt_occ)
table(id.by.mob.juv$agg_occ)
#Collinear X variables
ggpairs(id.by.mob.juv[,c(11,14:15,19,21,23,25,28)])
#Create juvenile dataset
head(id.by.mob.juv[,c(1:10,11,14:15,18,20,23,25,28)])
id.by.mob.juv.all <- id.by.mob.juv[,c(1:10,11,14:15,18,20,23,25,28)] #1170
#Fix outliers in AI - hk & sd littermates, wr & mh littermates
id.by.mob.juv.all[id.by.mob.juv.all$hyena == "hk" & id.by.mob.juv.all$mobID == 336,]$mob.ai.avg <- NA
id.by.mob.juv.all[id.by.mob.juv.all$hyena == "sd" & id.by.mob.juv.all$mobID == 336,]$mob.ai.avg <- NA
id.by.mob.juv.all[id.by.mob.juv.all$hyena == "wr" & id.by.mob.juv.all$mobID == 205,]$mob.ai.avg <- NA
id.by.mob.juv.all[id.by.mob.juv.all$hyena == "mh" & id.by.mob.juv.all$mobID == 205,]$mob.ai.avg <- NA
########## 16.3 Additive model ##########
##### Create dataset #####
head(id.by.mob.juv.all[,c(11:18)])
ggcorr(id.by.mob.juv.all[,c(11:18)], label = T)
# stan.rank and present.prop.rank.higher are too correlated - drop present.prop.rank.higher
head(id.by.mob.juv.all[,c(11:17)])
id.by.mob.juv.tmp <- filter(id.by.mob.juv.all, complete.cases(id.by.mob.juv.all[,c(11:17)])) #n=1139
id.by.mob.juv.tmp$stan.age <- as.numeric(scale(id.by.mob.juv.tmp$age, center = T, scale = T))
id.by.mob.juv.tmp$stan.rank.og <- id.by.mob.juv.tmp$stan.rank
id.by.mob.juv.tmp$stan.rank <- as.numeric(scale(id.by.mob.juv.tmp$stan.rank.og, center = T, scale = T))
id.by.mob.juv.tmp$stan.prop.mob.related <- as.numeric(scale(id.by.mob.juv.tmp$prop.mob.related,
center = T, scale = T))
id.by.mob.juv.tmp$stan.mob.ai.avg <- as.numeric(scale(id.by.mob.juv.tmp$mob.ai.avg,
center = T, scale = T))
summary(id.by.mob.juv.tmp)
ggcorr(id.by.mob.juv.tmp[,c(11:17)], label = T) #all below 0.7
##### Global model - additive #####
#Full model
mod.juv <- glmmTMB(mobber ~ stan.age + stan.rank + sex + grt_occ +
stan.prop.mob.related + stan.mob.ai.avg + (1|session/mobID) + (1|hyena),
data = id.by.mob.juv.tmp, family = binomial(link = 'logit'))
check_collinearity(mod.juv) #all below 3
check_model(mod.juv)
#Dredge
options(na.action = "na.fail")
mod.juv <- glmmTMB(mobber ~ stan.age + stan.rank + sex + grt_occ +
stan.prop.mob.related + stan.mob.ai.avg + (1|session/mobID) + (1|hyena),
data = id.by.mob.juv.tmp, family = binomial(link = 'logit'))
results.juv <- dredge(mod.juv)
get.models(subset(results.juv, delta == 0), subset = T)
# mobber ~ sex + stan.age + (1 | session/mobID) + (1 | hyena)
importance(subset(results.juv, delta <= 6 & !nested(.)))
# cond(stan.age) cond(sex)
# Sum of weights: 1.00 0.53
# N containing models: 2 1
options(na.action = "na.omit")
########## 16.4 Interactive model ##########
##### Global model - interactive #####
#Full model
mod.juv <- glmmTMB(mobber ~ stan.age + stan.rank + sex + grt_occ +
stan.prop.mob.related + stan.mob.ai.avg +
stan.age * stan.rank +
stan.age * sex +
stan.rank * sex + (1|session/mobID) + (1|hyena),
data = id.by.mob.juv.tmp, family = binomial(link = 'logit'))
check_collinearity(mod.juv) #all below 3
check_model(mod.juv)
#Dredge
options(na.action = "na.fail")
mod.juv <- glmmTMB(mobber ~ stan.age + stan.rank + sex + grt_occ +
stan.prop.mob.related + stan.mob.ai.avg +
stan.age * stan.rank +
stan.age * sex +
stan.rank * sex + (1|session/mobID) + (1|hyena),
data = id.by.mob.juv.tmp, family = binomial(link = 'logit'))
results.juv <- dredge(mod.juv)
get.models(subset(results.juv, delta == 0), subset = T)
# mobber ~ grt_occ + sex + stan.age + (1 | session/mobID) + (1 | hyena) + sex:stan.age
importance(subset(results.juv, delta <= 6 & !nested(.)))
# cond(stan.age) cond(sex) cond(sex:stan.age) cond(grt_occ)
# Sum of weights: 1.00 0.87 0.72 0.39
# N containing models: 4 3 2 1
options(na.action = "na.omit")
########## 16.5 Final model and plots ##########
head(id.by.mob.juv.all[,c(11:12,14)])
id.by.mob.juv.final <- filter(id.by.mob.juv.all, complete.cases(id.by.mob.juv.all[,c(11:12,14)])) #n=1153
id.by.mob.juv.final$stan.age <- as.numeric(scale(id.by.mob.juv.final$age, center = T, scale = T))
summary(id.by.mob.juv.final)
##### Best model #####
mod.juv <- glmmTMB(mobber ~ grt_occ + sex + stan.age + (1 | session/mobID) + (1 | hyena) + sex:stan.age,
data = id.by.mob.juv.final, family = binomial(link = 'logit'))
summary(mod.juv)
# Estimate Std. Error z value Pr(>|z|)
# grt_occTRUE 1.1673 0.7618 1.532 0.1255
# sexm -0.3550 0.3894 -0.912 0.3619
# stan.age 1.7483 0.3335 5.243 1.58e-07 ***
# sexm:stan.age -0.7127 0.4303 -1.656 0.0977 .
check_collinearity(mod.juv) #all below 3
check_model(mod.juv)
simulationOutput <- simulateResiduals(fittedModel = mod.juv, n = 250)
plot(simulationOutput)
model_performance(mod.juv)
# AIC | AICc | BIC | R2 (cond.) | R2 (marg.) | ICC | RMSE | Sigma | Log_loss | Score_log | Score_spherical
# ----------------------------------------------------------------------------------------------------------------------
# 795.042 | 795.168 | 835.443 | 0.722 | 0.178 | 0.662 | 0.227 | 1.000 | 0.178 | -Inf | 0.015
##### Plot models #####
id.by.mob.juv.final$mobber <- as.factor(id.by.mob.juv.final$mobber)
id.by.mob.juv.final$sex <- as.factor(id.by.mob.juv.final$sex)
id.by.mob.juv.final$grt_occ <- as.factor(id.by.mob.juv.final$grt_occ)
mod.juv <- glmmTMB(mobber ~ grt_occ + sex + stan.age + (1 | session/mobID) + (1 | hyena) + sex:stan.age,
data = id.by.mob.juv.final, family = binomial(link = 'logit'))
summary(mod.juv)
#Sex
dat.sex <- ggeffects::ggpredict(mod.juv, type = "fe", terms = c("sex"), full.data = FALSE)
dat.sex$x <- factor(dat.sex$x, labels = c("female", "male"))
plot.sex <- ggplot(dat.sex, aes(x = x, y = predicted)) +
geom_point(size = 5) +
geom_line(data = data.frame(x = c(dat.sex$x, dat.sex$x),
predicted = c(dat.sex$predicted, dat.sex$predicted),
ci = c(dat.sex$conf.low, dat.sex$conf.high)),
aes(x = x, y = ci, group= x), size = 1.5)+
xlab('Sex') +
ylab("Predicted probability of mobbing")+
theme_classic() +
theme(legend.title = element_blank(), legend.text = element_text(size = 12),
axis.title = element_text(size = 20), axis.text = element_text(size = 16))+
ylim(c(0,1))
pdf('16.plot.juv.sex.pdf', width = 7, height = 5)
plot.sex
dev.off()
#Age
summary(id.by.mob.juv.final$age)
dat.age <- ggeffects::ggpredict(mod.juv, terms = c("stan.age [all]"), full.data = FALSE)
plot.age <- ggplot(dat.age, aes(x = dat.age$x*sd(id.by.mob.juv.final$age, na.rm = T) +
mean(id.by.mob.juv.final$age, na.rm = T),
y = predicted)) +
geom_line(size = 1) +
geom_ribbon(aes(x = dat.age$x*sd(id.by.mob.juv.final$age, na.rm = T) + mean(id.by.mob.juv.final$age, na.rm = T),
ymin = conf.low, ymax = conf.high), alpha = 0.2, inherit.aes = FALSE) +
xlab('Age') +
ylab("Predicted probability of mobbing")+
theme_classic() +
theme(legend.title = element_blank(), legend.text = element_text(size = 12),
axis.title = element_text(size = 20), axis.text = element_text(size = 16))+
ylim(c(0,1))
pdf('16.plot.juv.age.pdf', width = 7, height = 5)
plot.age
dev.off()
#Greetings
dat.grt <- ggeffects::ggpredict(mod.juv, type = "fe", terms = c("grt_occ"), full.data = FALSE)
plot.grt <- ggplot(dat.grt, aes(x = x, y = predicted)) +
geom_point(size = 5) +
geom_line(data = data.frame(x = c(dat.grt$x, dat.grt$x),
predicted = c(dat.grt$predicted, dat.grt$predicted),
ci = c(dat.grt$conf.low, dat.grt$conf.high)),
aes(x = x, y = ci, group= x), size = 1.5)+
xlab('Greeted') +
ylab("Predicted probability of mobbing") +
theme_classic() +
theme(legend.title = element_blank(), legend.text = element_text(size = 12),
axis.title = element_text(size = 20), axis.text = element_text(size = 16),
axis.title.y = element_blank()) +
ylim(c(0,1))
pdf('16.plot.juv.grt.pdf', width = 7, height = 5)
plot.grt
dev.off()
#Plot of single variables
layout <- "A#
BC"
plots <- plot.age + plot.sex + plot.grt + plot_layout(design = layout)
pdf('16.plots.all.pdf', width = 7, height = 5)
plots
dev.off()
#Interaction between age and sex
summary(id.by.mob.juv.final$age)
dat.intx.sex.age <- ggeffects::ggpredict(mod.juv, terms = c("stan.age [all]", "sex"), full.data = FALSE)
dat.intx.sex.age$group <- factor(dat.intx.sex.age$group, labels = c("female", "male"))
plot.sex.age <- ggplot(dat.intx.sex.age, aes(x = dat.intx.sex.age$x*sd(id.by.mob.juv.final$age, na.rm = T) +
mean(id.by.mob.juv.final$age, na.rm = T),
y = predicted, color = group, fill = group)) +
geom_ribbon(aes(x = dat.intx.sex.age$x*sd(id.by.mob.juv.final$age, na.rm = T) +
mean(id.by.mob.juv.final$age, na.rm = T),
ymin = conf.low, ymax = conf.high, fill = dat.intx.sex.age$group), alpha = 0.2, inherit.aes = FALSE) +
geom_line(size = 1) +
xlab('Age') +
ylab("Predicted probability of mobbing")+
theme_classic() +
theme(legend.title = element_text(size = 16), legend.text = element_text(size = 12),
axis.title = element_text(size = 20), axis.text = element_text(size = 16),
legend.position = "none") + #"top"
scale_color_manual(values = viridis_3, name = "Sex", labels = c("Female", "Male")) +
scale_fill_manual(values = viridis_3, name = "Sex", labels = c("Female", "Male")) +
ylim(c(0,1))
pdf('16.plot.juv.age.sex.pdf', width = 7, height = 5)
plot.sex.age
dev.off()
#Model
mod.juv <- glmmTMB(mobber ~ stan.age + sex + grt_occ + (1 | session/mobID) + (1 | hyena) + sex:stan.age,
data = id.by.mob.juv.final, family = binomial(link = 'logit'))
summary(mod.juv)
set_theme(base = theme_classic(), axis.textcolor = "black", axis.title.color = "black",
axis.textsize.y = 1.5, axis.textsize.x = 1.2, axis.title.size = 1.7)
plot.model <- plot_model(mod.juv, type = "est", transform = "exp",
axis.labels = c("Age x Sex", "Greeted [TRUE]",
"Sex [male]", "Age"),
vline.color = "black", title = "", dot.size = 4.5, line.size = 1.5,
show.values = TRUE, show.p = TRUE, digits = 2, value.offset = 0.3,
value.size = 6, colors = viridis_2, axis.lim = c(0.01,100))
pdf('16.plot.juv.model.pdf', width = 7, height = 5)
plot.model
dev.off()
#Table
sjPlot::tab_model(mod.juv, show.se = T, show.ci = 0.95, show.re.var = F, show.intercept = F,
pred.labels = c("Age", "Sex [male]", "Greeted [TRUE]",
"Age x Sex"),
dv.labels = c("Probability of mobbing participation"),
string.se = "SE", transform = "exp", file = "16.table_juv.doc")
#Calculate CIs using the likelihood profile
tidy.mod.juv <- broom.mixed::tidy(mod.juv, conf.method = "profile",
conf.int = T, conf.level = 0.95, exponentiate = T)
tidy.mod.juv
# effect component group term estimate std.error statistic p.value conf.low conf.high
# 2 fixed cond NA stan.age 5.75 1.92 5.24 1.58e-7 3.16 12.0
# 3 fixed cond NA sexm 0.701 0.273 -0.912 3.62e-1 0.317 1.51
# 4 fixed cond NA grt_occTRUE 3.21 2.45 1.53 1.25e-1 0.710 14.7
# 5 fixed cond NA stan.age:sexm 0.490 0.211 -1.66 9.77e-2 0.202 1.13
# 6 ran_pars cond mobID:session sd__(Intercept) 1.08 NA NA NA -0.803 0.610
# 7 ran_pars cond session sd__(Intercept) 1.88 NA NA NA 0.151 1.03
# 8 ran_pars cond hyena sd__(Intercept) 1.32 NA NA NA -0.188 0.703