-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathsenate_model.R
160 lines (116 loc) · 4.79 KB
/
senate_model.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
# the model ------------------------------------------------------------
library(lubridate)
library(tidyverse)
library(ggmcmc)
library(rjags)
library(civis)
#dat <- read_civis("cody.polling_cleaned_1102") %>% as.tibble()
model <- "
model {
### Likelihood
for (i in 1:length(twoway)) {
twoway[i] ~ dnorm(mu[i], prec[i])
# yhat[i] ~ dnorm(mu[i], 5)
mu[i] = xi[week_adj[i], state_num[i]] + delta[pollster_num[i]] + theta[univ_num[i]] + xi_week[week_adj[i]] + gamma[state_num[i]]
}
## predictions
for (j in 1:length(gammas)) {
for (t in 1:length(xi_week)) {
pred[t,j] = xi[t, j] + xi_week[t] + gamma[j]
}
}
## time priors:
for (j in 1:length(gammas)) {
for (t in 2:length(xi_week)) {
xi[t,j]~dnorm(xi[t-1, state_num[j]], tau_stateweek[j])
}
xi[1, j]~dnorm(xi1_prior[j] + state_bias[j], tau_stateweek[j])
xi1_prior[j] ~ dunif(.01, .5)
mix_stateweek[j] ~ dcat(mix_prior_prob[])
omega_stateweek[1, j] ~ dunif(0, 0.04)
omega_stateweek[2, j] ~ dunif(0, 0.004)
tau_stateweek[j] <- 1/pow(omega_stateweek[mix, j], 2)
}
for (t in 2:length(xi_week)) {
xi_week[t]~dnorm(xi_week[t-1], tau_week)
}
xi_week[1]~dunif(.01, .5)
mix ~ dcat(mix_prior_prob[])
mix_prior_prob[1] <- .02
mix_prior_prob[2] <- .98
omega[1] ~ dunif(0, 0.035)
omega[2] ~ dunif(0, 0.0035)
tau_week <- 1/pow(omega[mix],2)
## Pollster priors:
for (i in 1:length(deltas)) {
delta[i] ~ dnorm(pollster_bias[i], tau_delta)
}
omega_delta ~ dunif(0, .05) #.00560)
tau_delta = 1/pow(omega_delta, 2)
## state priors
for (i in 1:length(gammas)) {
gamma[i] ~ dnorm(state_bias[i], tau_gamma)
}
omega_gamma ~ dunif(0, .1)
tau_gamma = 1/pow(omega_gamma, 2)
## Universe priors:
for (i in 1:length(thetas)) {
theta[i] ~ dnorm(theta_mu_prior[i], 1/theta_var_prior[i])
}
}
"
# data and running model --------------------------------------------------
## extracting pollster and state info for JAGS
pollster_leans <- dat %>% select(pollster, pollster_num, bias, var_upper) %>% unique
state_leans <- dat %>% select(state, state_lean, state_num) %>% unique %>% arrange(state_num) %>% select(state_lean)
## making the jags object
data = with(dat,
list(
twoway = twoway,
state_num = state_num,
week_adj = week_adj,
# month_adj = month_adj,
pollster_num = pollster_num,
univ_num = univ_num,
pollster_bias = pollster_leans$bias,
state_bias = state_leans$state_lean,
# pollster_var_lower = var_lower,
# pollster_var_upper = var_upper,
prec = 1 / ((twoway * (1 - twoway)) / n_size),
deltas = rep(NA, length(pollster %>% unique)),
thetas = rep(NA, length(univ %>% unique)),
gammas = rep(NA, length(state_num %>% unique)),
xi = matrix(data = NA, nrow = length(rep(NA, max(length(unique(week_adj)), max(week_adj)))), ncol = length(state_num %>% unique)),
pred = matrix(data = NA, nrow = length(rep(NA, max(length(unique(week_adj)), max(week_adj)))), ncol = length(state_num %>% unique)),
xi_week = rep(NA, max(length(unique(week_adj)), max(week_adj))),
# xi_month = rep(NA, max(length(unique(month_adj)), max(month_adj))),
# theta_mu_prior = c(0, 0, 0),
# theta_var_prior = c(10, 10, 10)
# theta_mu_prior = c(.00925, .0131, .0386),
theta_mu_prior = c(.0386, -.01, .026),
theta_var_prior = c(0.00000676, 0.00000827, .000038)
#theta_var_prior = c(1000,1000,1000)
))
#lapply(data, len
# run model ---------------------------------------------------------------
mod <- jags.model(textConnection(model),
data = data,
n.chains = 3)
samp <- coda.samples(mod,
variable.names = c("xi", "delta", "theta", "xi_week", "gamma",
# "tau_gamma", "tau_week", "tau_stateweek",
"pred",
"omega", "omega_stateweek",
"mix", "mix_stateweek"),
burnin = 10000,
n.iter = 1000000,
adapt = 1000,
thin = 1000
# burnin = 10000,
# n.iter = 10000,
# adapt = 1000,
# thin = 10
)
s <- ggs(samp)
#s %>% write.csv("1031_1m_iter.csv")
#s %>% write_civis("cody.senate_model_output") ## note this takes a really long time for the big simulatoins - I'd recommend keeping this all local given we're updating this every day.