forked from bjbarrett/multinomial_freq_beyond_two
-
Notifications
You must be signed in to change notification settings - Fork 0
/
02_multinomial_sim_plots.R
204 lines (178 loc) · 7.4 KB
/
02_multinomial_sim_plots.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
library(rethinking)
library(RColorBrewer)
#load("~/Documents/multinomial_freq_beyond_two/100simsK2345_popsame.rdata")
load("~/Documents/multinomial_freq_beyond_two/100simsK2345_differentdistforagentspersim.rdata")
CalcEntropy<- function(x){
out = -sum(ifelse(x==0, 0 , x*log(x)) )
return(out)
}
drops <- c("choice","n_i","obs")
##2
d2 <- dsim2[ , !(names(dsim2) %in% drops)]
d2 <- d2[!duplicated(d2), ]
d2$entropy <- as.vector(apply(d2[,1:2] ,1, CalcEntropy))
d2$post_f_med <- as.vector(apply(post2$f ,2, median))
d2$loss <- abs(d2$f-d2$post_f_med)
##3
d3 <- dsim3[ , !(names(dsim3) %in% drops)]
d3 <- d3[!duplicated(d3), ]
d3$entropy <- as.vector(apply(d3[,1:3] ,1, CalcEntropy))
d3$post_f_med <- as.vector(apply(post3$f ,2, median))
d3$loss <- abs(d3$f-d3$post_f_med)
## 4
d4 <- dsim4[ , !(names(dsim4) %in% drops)]
d4 <- d4[!duplicated(d4), ]
d4$entropy <- as.vector(apply(d4[,1:4] ,1, CalcEntropy))
d4$post_f_med <- as.vector(apply(post4$f ,2, median))
d4$loss <- abs(d4$f-d4$post_f_med)
## 5
d5 <- dsim5[ , !(names(dsim5) %in% drops)]
d5 <- d5[!duplicated(d5), ]
d5$entropy <- as.vector(apply(d5[,1:5] ,1, CalcEntropy))
d5$post_f_med <- as.vector(apply(post5$f ,2, median))
d5$loss <- abs(d5$f-d5$post_f_med)
d <- rbind(d2,d3,d4,d5)
post_f <- cbind(post2$log_f, post3$log_f , post4$log_f , post5$log_f)
postposi <- function(x){sum( x > 0 ) / 2000 }
d$perc_positive <- apply( post_f , 2 , postposi )
d$sim_sd <- apply( post_f , 2 , sd )
d$med_log_f <- apply( post_f , 2 , median )
N <- sort(unique(d$n)) ## pop size vector from data
F <-sort(unique(d$f)) ## strength of frequency dependence from data
K <- sort(unique(d$k)) ## number of options from data
nsims <- 100
pal_names=c("Oranges","Greens","Blues","Purples")
for (f_i in 1:length(F)){
pdf(file = paste("posts_sims_f_",F[f_i],".pdf") , width = 8, height = 8) # The height of the plot in inches
par( mfrow = c( length(N) , length(K) ) )
par( oma=c(4,1,1,0) +.2)
par(mar=c(0,0,0,0)+.1)
for (n_i in 1:length(N)){
for (k_i in 1:length(K)){
mypalette <- brewer.pal(9 , pal_names[k_i])[4:9]
index <- which(d$f==F[f_i] & d$k==K[k_i] & d$n==N[n_i])
dens( post_f[,index] , col=mypalette[n_i] , xlim=c(-3,3) ,ylim=c(0,7) , xaxt='n' , yaxt='n') #post of all sims
abline(v=log(F[f_i]) , lw=2 , col=mypalette[n_i]) #line at true value
abline(v=0 , lty=3) #line at conform grenze
seq_l <- seq(from=0 , to=6 , length=nsims) #vertical range to plot hpdi segs
for (i in index) dens(post_f[,i] , add=TRUE , col=col.alpha(mypalette[n_i] , alpha=0.15))
dens(rnorm(1e7 , mean=0 , sd=1) , lty=2 , add=TRUE)
axis(1, at=c(-4:4) , cex.axis=0.8 , tck=0.01 , labels=FALSE )
axis(1, at=c(-4:4) , cex.axis=0.8 , tck=-0.01 , labels=FALSE )
if(n_i==length(N)){
axis(1, at=c(-4:4), labels=c(-4:4) , cex.axis=0.8 )
}
if(n_i==1){
mtext(paste("k =",K[k_i]) , line=0 , cex.main=1, bg="white")
}
if(k_i==1){
mtext(paste("n=",N[n_i]) , line=0 , cex.main=1, bg="white" , side=2)
}
}
}
mtext(side=1 ,"posterior density of log(f)", outer = TRUE, line=2.2 )
dev.off()
}
#plot(loss ~ entropy , data=d[d$f==3 & d$n==25 & d$k==4,] , col=mypalette[2])
##alternative with sd bars
for (f_i in 1:length(F)){
pdf(file = paste("posts_sims_ci_f_",F[f_i],".pdf") , width = 8, height = 8) # The height of the plot in inches
par( mfrow = c( length(N) , length(K) ) )
par( oma=c(4,1,1,0) +.2)
par(mar=c(0,0,0,0)+.1)
for (n_i in 1:length(N)){
for (k_i in 1:length(K)){
mypalette <- brewer.pal(9 , pal_names[k_i])[4:9]
index <- which(d$f==F[f_i] & d$k==K[k_i] & d$n==N[n_i])
plot( 0 ,0, col="white" , xlim=c(-3,3) ,ylim=c(0,4) , xaxt='n' , yaxt='n') #post of all sims
abline(v=log(F[f_i]) , lw=2 , col=mypalette[n_i]) #line at true value
abline(v=0 , lty=3) #line at conform grenze
seq_l <- seq(from=0 , to=4 , length=nsims) #vertical range to plot hpdi segs
f_med <- apply(post_f[,index] , 2 , median) #actual medians
f_hpdi <- apply(post_f[,index] , 2 , HPDI, prob=30*pi/100) #actual medians
for(i in 1:nsims){
segments( x0=f_hpdi[,order(f_med)[i]][1] , y0=seq_l[i] ,
x1=f_hpdi[,order(f_med)[i]][2], y1=seq_l[i] , col=col.alpha("black" , alpha=.7) , lty=1 , lw=.5)
}
points(f_med[order(f_med)] , seq_l , cex=0.25 , col=c(mypalette[n_i]) , pch=5)
axis(1, at=c(-4:4) , cex.axis=0.8 , tck=0.01 , labels=FALSE )
axis(1, at=c(-4:4) , cex.axis=0.8 , tck=-0.01 , labels=FALSE )
if(n_i==length(N)){
axis(1, at=c(-4:4), labels=c(-4:4) , cex.axis=0.8 )
}
if(n_i==1){
mtext(paste("k =",K[k_i]) , line=0 , cex.main=1, bg="white")
}
if(k_i==1){
mtext(paste("n=",N[n_i]) , line=0 , cex.main=1, bg="white" , side=2)
}
}
}
mtext(side=1 ,"predicted value of log(f)", outer = TRUE, line=2.2 )
dev.off()
}
#
# for (f_i in 3:length(F)){
# for (n_i in 1:length(N)){
# dens( c(0,1) , col="white" , xlim=c(-2,3) ,ylim=c(0,3.5) , xaxt='n' , yaxt='n' ) #post of all sims
# for (k_i in 1:length(K)){
# index <- which(d$f==F[f_i] & d$k==K[k_i] & d$n==N[n_i])
# textz <- paste( n =N[n_i]) #title of plots
# dens( post_f[,index] , col=mypalette[k_i] , add=TRUE) #post of all sims
# title(main=textz , line=-.7 , cex.main=0.8 , bg="white")
# abline(v=log(F[f_i]) , lw=2 , col=mypalette[k_i]) #line at true value
# #abline(v=0 , lty=3) #line at conform grenze
# #dens(rnorm(1e7 , mean=0 , sd=1) , lty=2 , add=TRUE)
# axis(1, at=c(-4:4), labels=c(-4:4) , cex.axis=0.8 )
#
# }
# }
# mtext(side=1 ,"posterior density of log(f)", outer = TRUE, line=2.2 )
# }
#plot % of posterior mass > 0 for each sim at each sample size, colors k , panels are samples sizes, point types correspond with type
# for (n_i in 1:length(N)){
# plot(0,0 , xlim=c(1,100) , ylim=c(0,1) , col="white" , xlab="simulation number" , ylab="propotion of positive posterior samples of log(f)" )
# for(f_i in 1:length(F)){
# for (k_i in 1:length(K)){
# index <- which(d$f==F[f_i] & d$k==K[k_i] & d$n==N[n_i])
# yy <- d$perc_positive[index]
# points(1:100 , yy[order(yy)] , cex=0.5 , col=c(mypalette[k_i]) , pch=f_i+15 , type="b")
#
# }
# }
# textz <- paste("n =",N[n_i] ) #title of plots
# title(main=textz , line=0.5 , cex.main=1 , bg="white")
# }
#
# str(d)
#
# plot(d$loss , col=mypalette[d$k] )
#
# plot(perc_positive~med_log_f, data=d[d$f==3 & d$n==250,] , col=col.alpha(mypalette[d$k[d$f==3 & d$n==250]] , alpha=0.2) )
# postposi <- function(x){sum( x > 0 ) / 2000 }
#
#
# d$perc_positive <- apply( post_f , 2 , postposi )
# d$sim_sd <- apply( post_f , 2 , sd )
# d$med_log_f <- apply( post_f , 2 , median )
#
#
# for (n_i in 1:length(N)){
# plot(0,0 , xlim=c(1,100) , ylim=c(0,1) , col="white" , xlab="simulation number" , ylab="propotion of positive posterior samples of log(f)" )
# for(f_i in 1:length(F)){
# for (k_i in 1:length(K)){
# index <- which(d$f==F[f_i] & d$k==K[k_i] & d$n==N[n_i])
# yy <- d$perc_positive[index]
# points(1:100 , yy[order(yy)] , cex=0.5 , col=c(mypalette[k_i]) , pch=f_i+15 , type="b")
#
# }
# }
# textz <- paste("n =",N[n_i] ) #title of plots
# title(main=textz , line=0.5 , cex.main=1 , bg="white")
# }
#
# str(d)
#
# plot(d$loss , col=mypalette[d$k])
#
# plot(perc_positive~med_log_f, data=d[d$f==3 & d$n==250,] , col=mypalette[d$k[d$f==3 & d$n==250]])