-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy path03_modelagnostic_analysis.R
executable file
·259 lines (208 loc) · 10.9 KB
/
03_modelagnostic_analysis.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
rm(list=ls())
load('03_data/02_aggregated_data/wurs.Rdata')
load('03_data/02_aggregated_data/asrs.Rdata')
load('03_data/02_aggregated_data/bis.Rdata')
load('03_data/02_aggregated_data/bisbas.Rdata')
load('03_data/02_aggregated_data/task_clean.Rdata')
load('03_data/02_aggregated_data/open_questions.Rdata')
library(dplyr)
library(tidyr)
library(lme4)
library(ggplot2)
library(viridis)
# subclinical adhd score --------------------------------------------------
self.report<-data.frame(subj=asrs$subj,asrs=asrs$asrs6,asrs2=asrs$asrs.partb,wurs=wurs$wurs,bis=bis$bis)
cor(self.report)
self.report$compound<-rowMeans(self.report[,-1])
plot(self.report$compound)
hist(self.report$compound)
# bis bas questionnaire --------------------------------------------------
plot(bisbas$BIS_1)
hist(bisbas$BIS_1)
plot(bisbas$BAS_Drive)
hist(bisbas$BAS_Drive)
plot(bisbas$BAS_Fun_Seeking)
hist(bisbas$BAS_Fun_Seeking)
plot(bisbas$BAS_Reward_Responsiveness)
hist(bisbas$BAS_Reward_Responsiveness)
# model-agnostic effects --------------------------------------------------
#reward effect
df_task_reward_effect_bandit<-df_task_clean%>%group_by(subj,reward_oneback)%>%summarise(pStay_badnit1=mean(stay1_bandit),pStay_bandit2=mean(stay2_bandit))
df_task_reward_effect_bandit_group<-df_task_clean%>%group_by(reward_oneback)%>%summarise(pStay_badnit1=mean(stay1_bandit),pStay_bandit2=mean(stay2_bandit))
df_task_reward_resp_bandit<-df_task_clean%>%group_by(reward_oneback,key1_oneback,key2_oneback)%>%summarise(pStay_badnit1=mean(stay1_bandit),pStay_bandit2=mean(stay2_bandit))
plot(df_task_reward_resp_bandit$pStay_badnit1)
plot(df_task_reward_resp_bandit$pStay_bandit2)
df_task_reward_effect_resp<-df_task_clean%>%group_by(reward_oneback,key1_oneback,key2_oneback)%>%summarise(pStay_key1=mean(stay1_key),pStay_key2=mean(stay2_key))
plot(df_task_reward_effect_resp$pStay_key2)
#stay probability for second bandit as function of previous outcome and previous second stage response
#and the probability to repeat second stage response as a function of previous outcome and previous second stage response
df_task_reward_resp_bandit_2<-df_task_clean%>%group_by(reward_oneback,key2_oneback)%>%summarise(pStay_bandit2=mean(stay2_bandit),pStay_key2=mean(stay2_key))
plot(df_task_reward_resp_bandit_2$pStay_bandit2)
plot(df_task_reward_resp_bandit_2$pStay_key2)
df_task_reward_resp_bandit_1<-df_task_clean%>%group_by(reward_oneback,key1_oneback)%>%summarise(pStay_bandit1=mean(stay1_bandit),pStay_key1=mean(stay1_key))
plot(df_task_reward_resp_bandit_1$pStay_bandit1)
plot(df_task_reward_resp_bandit_1$pStay_key1)
names(df_task_clean)
#1st stage repetition
#the reward effect for sequences of : Ng-Ng, Ng-Go, Ng-Go, Go-Go
df_task_clean%>%
filter()%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay1_bandit))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)
#2nd stage repetition
#the reward effect for sequences of : Ng-Ng, Ng-Go, Ng-Go, Go-Go
#only for trails that the previous second stage was the same. (meaning, same first stage choice as the previous trail)
df_task_clean%>%
filter(stay2_state)%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)
df_task_clean%>%
filter(stay2_state)%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(subj,reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)%>%
ggplot(aes(x=factor(key1_oneback), y=reward_effect, color=factor(key2_oneback))) +
geom_boxplot() +
scale_fill_viridis(discrete = TRUE, alpha=0.6) +
geom_jitter(size=0.4, position=position_jitter(0.3)) +
ggtitle("2nd stage repetition") +
xlab("key1_oneback")+ ylab("reward_effect")
#hierarchical logistic regression model ???
m <- glmer(stay2_bandit ~ 1+reward_oneback+reward_oneback:key1_oneback+reward_oneback:key1_oneback+
(1+reward_oneback+reward_oneback:key1_oneback+reward_oneback:key1_oneback+ | subj),
data = df_task_clean,
family = binomial, control = glmerControl(optimizer = "bobyqa"),
nAGQ = 0)
summay(m)
#stay probability after rewarded or unrewarded trails as a factor of the responses at the previos first and second stages.
#only for trails that the previos second stage was the same. (meaning, same first stage choice as the previos trail)
df_seq <- df_task_clean%>%
filter(stay2_state)%>%
group_by(reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))
plot(df_seq$pStay)
#separately for each subject
df_task_clean%>%
filter(stay2_state)%>%
group_by(subj,reward_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))%>%
ggplot(aes(x=factor(key2_oneback), y=pStay,color=factor(reward_oneback))) +
geom_boxplot()
####
#2nd stage repetition
#the reward effect for sequences of : Ng-Ng, Ng-Go, Ng-Go, Go-Go
df_task_clean%>%
filter(stay2_state)%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)
#hierarchical logistic regression model
#interaction between reward effect and response in the first and second stages at the previos trail
m <- glmer(stay2_bandit ~ 1+reward_oneback*key1_oneback*key2_oneback +
(1+reward_oneback*key1_oneback*key2_oneback | subj),
data = df_task_clean,
family = binomial, control = glmerControl(optimizer = "bobyqa"),
nAGQ = 0)
summary(m)
x= coef(m)$subj[,'reward_oneback']
y= self.report$compound
fit <- lm(y ~ x) ## polynomial of degree 3
plot(x, y,xlab = 'reward_oneback',ylab = 'self reports score') ## scatter plot (colour: black)
x0 <- seq(min(x), max(x), length = 100) ## prediction grid
y0 <- predict.lm(fit, newdata = list(x = x0)) ## predicted values
lines(x0, y0, col = 2) ## add regression curve (colour: red)
x= coef(m)$subj[,'reward_oneback:key1_oneback']
y= self.report$compound
fit <- lm(y ~ x) ## polynomial of degree 3
plot(x, y,xlab = 'reward_oneback:key1_oneback',ylab = 'self reports score') ## scatter plot (colour: black)
x0 <- seq(min(x), max(x), length = 100) ## prediction grid
y0 <- predict.lm(fit, newdata = list(x = x0)) ## predicted values
lines(x0, y0, col = 3) ## add regression curve (colour: green)
x= coef(m)$subj[,'reward_oneback:key2_oneback']
y= self.report$compound
fit <- lm(y ~ x) ## polynomial of degree 3
plot(x, y,xlab = 'reward_oneback:key2_oneback',ylab = 'self reports score') ## scatter plot (colour: black)
x0 <- seq(min(x), max(x), length = 100) ## prediction grid
y0 <- predict.lm(fit, newdata = list(x = x0)) ## predicted values
lines(x0, y0, col = 5) ## add regression curve (colour: blue)
x= coef(m)$subj[,'reward_oneback:key1_oneback:key2_oneback']
y= self.report$compound
fit <- lm(y ~ x) ## polynomial of degree 3
plot(x, y,xlab = 'reward_oneback:key1_oneback:key2_oneback',ylab = 'self reports score') ## scatter plot (colour: black)
x0 <- seq(min(x), max(x), length = 100) ## prediction grid
y0 <- predict.lm(fit, newdata = list(x = x0)) ## predicted values
lines(x0, y0, col = 6) ## add regression curve (colour: purple)
####
df_1st_choice <- df_task_clean%>%
group_by(reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay1_bandit))%>%
ggplot(aes(x=factor(key2_oneback), y=pStay,color=factor(reward_oneback))) +
geom_boxplot()
plot(df_1st_choice$pStay)
#check corr with subclinical adhd
effects<-
df_task_clean%>%
filter(stay2_state)%>%
group_by(subj,reward_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))%>%
pivot_wider(names_from = c('reward_oneback','key2_oneback'),values_from='pStay')
plot(self.report$compound,unlist(c(effects[,4]-effects[,2])))
plot(self.report$compound,unlist(c(effects[,5]-effects[,3])))
#check corr with bisbas score
# effects of reward on key-stay -------------------------------------------
#KEY2
#how likely it is to repeat the second stage response as a function of : the orevios first and second stage action and whether the trail was rewarded or not
h<-df_task_clean%>%
filter(stay2_state==F)%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(reward_oneback,key1_oneback,key2_oneback,key1)%>%
summarise(pStay=mean(stay2_key))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)
plot(h$reward_effect)
#hierarchical logistic regression model (NHT)
library(lme4)
m <- glmer(stay2_key ~ reward_oneback*key2_oneback*key1_oneback+(reward_oneback| subj), data = df_task_clean, family = binomial, control = glmerControl(optimizer = "bobyqa"), nAGQ = 0)
summary(m)
#bayesian hierarchical logistic regression
library(brms)
#m<-brm(stay2_key ~ reward_oneback*key2_oneback*key1_oneback + (reward_oneback*key2_oneback*key1_oneback|subj), data = df, family = bernoulli(link = "logit"),cores=20)
load('02_models/brms_stay2_key.Rdata')
summary(m)
plot(m, pars = c("reward_oneback"))
cor.test(self.report$compound,ranef(m)$subj[,,'reward_oneback'][,1])
# pStay bandit as a function of mapping -----------------------------------
names(df_task_clean)
df_task_clean%>%
filter(stay2_state==T)%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(stay2_mapping,reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_bandit))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)
library(lme4)
m <- glmer(stay2_bandit ~ reward_oneback*stay2_mapping+(reward_oneback| subj), data = df_task_clean[df_task_clean$stay2_state==T,], family = binomial, control = glmerControl(optimizer = "bobyqa"), nAGQ = 0)
summary(m)
##########
names(df_task_clean)
df_task_clean%>%
filter(stay2_state==F)%>%
filter(stay1_key==1)%>%
mutate(reward_oneback=factor(reward_oneback,levels=c(0,1),labels=c('unrewarded','rewarded')))%>%
group_by(reward_oneback,key1_oneback,key2_oneback)%>%
summarise(pStay=mean(stay2_key))%>%
pivot_wider(names_from = c('reward_oneback'),values_from=c('pStay'))%>%
mutate(reward_effect=rewarded-unrewarded)
library(lme4)
m <- glmer(stay2_bandit ~ reward_oneback*stay2_mapping+(reward_oneback| subj), data = df_task_clean[df_task_clean$stay2_state==T,], family = binomial, control = glmerControl(optimizer = "bobyqa"), nAGQ = 0)
summary(m)