Update code to include latest data and run for 2022 election

model-2pp/model-2pp.R

@@ -1,25 +1,26 @@
 #----------------set up----------------
 source("setup.R")
 rstan_options(auto_write = TRUE)
-options(mc.cores = 7)
+options(mc.cores = 4)
 
 ozpolls <- ozpolls %>%
   mutate(firm = ifelse(firm == "YouGov/Galaxy", "Galaxy", firm))
 
 
 #--------------Data preparation-------------------
 first_election <- as.Date("2007-11-24")
-next_election <- as.Date("2019-05-18")
+next_election <- as.Date("2022-05-21")
 firms_today <- ozpolls %>%
-  filter(election_year == 2019) %>%
+  filter(election_year == 2022) %>%
   distinct(firm) %>%
   pull(firm)
 
 # We don't have sample sizes except in the build up to 2016, so we'll just get
 # the average sample size for firms from them and assume that's what they stick to:
-sample_sizes <- ozpolls_2016 %>%
+sample_sizes <- ozpolls_2022 %>%
   mutate(firm = str_squish(gsub("\\(.+\\)", "", firm)),
          firm = ifelse(firm == "Morgan", "Roy Morgan", firm),
+         firm = ifelse(firm == "Newspoll-YouGov", "Newspoll", firm),
          firm = ifelse(firm == "ReachTel", "ReachTEL", firm)) %>%
   filter(firm != "Election result") %>%
   group_by(firm) %>%
@@ -32,7 +33,7 @@ sample_sizes <- sample_sizes  %>%
   mutate(ss = ifelse(is.na(ss), min(sample_sizes$ss, na.rm = TRUE), ss))
 
 all_firms <- ozpolls %>%
-  filter(!firm %in% c("Election result") &
+  filter(!firm %in% c("Election result", "ANU", "Resolve Strategic") &
            firm %in% firms_today) %>%
   pull(firm) %>%
   unique() %>%
@@ -71,7 +72,7 @@ one_pollster <- lapply(all_firms, function(x){
 
 # Put all the data into a single list we can pass to Stan:
 model_data <- list(
-  number_elections = 5,
+  number_elections = 6,
   election_days = election_days,
   election_results = election_results,
 
@@ -96,22 +97,22 @@ model_data <- list(
   y4_values = one_pollster[[4]]$intended_vote,
   y4_days = one_pollster[[4]]$n_days,
   y4_n = nrow(one_pollster[[4]]),
-  y4_se = one_pollster[[4]]$se[1],
+  y4_se = one_pollster[[4]]$se[1]
 
-  y5_values = one_pollster[[5]]$intended_vote,
-  y5_days = one_pollster[[5]]$n_days,
-  y5_n = nrow(one_pollster[[5]]),
-  y5_se = one_pollster[[5]]$se[1],
+  # y5_values = one_pollster[[5]]$intended_vote,
+  # y5_days = one_pollster[[5]]$n_days,
+  # y5_n = nrow(one_pollster[[5]]),
+  # y5_se = one_pollster[[5]]$se[1],
+  # 
+  # y6_values = one_pollster[[6]]$intended_vote,
+  # y6_days = one_pollster[[6]]$n_days,
+  # y6_n = nrow(one_pollster[[6]]),
+  # y6_se = one_pollster[[6]]$se[1]
 
-  y6_values = one_pollster[[6]]$intended_vote,
-  y6_days = one_pollster[[6]]$n_days,
-  y6_n = nrow(one_pollster[[6]]),
-  y6_se = one_pollster[[6]]$se[1],
-
-  y7_values = one_pollster[[7]]$intended_vote,
-  y7_days = one_pollster[[7]]$n_days,
-  y7_n = nrow(one_pollster[[7]]),
-  y7_se = one_pollster[[7]]$se[1]
+  # y7_values = one_pollster[[7]]$intended_vote,
+  # y7_days = one_pollster[[7]]$n_days,
+  # y7_n = nrow(one_pollster[[7]]),
+  # y7_se = one_pollster[[7]]$se[1]
 
   )
 

model-2pp/model-2pp.stan

@@ -29,26 +29,26 @@ data {
   int y4_days[y4_n];                     
   real y4_se;
 
-  int y5_n;
-  real y5_values[y5_n];       
-  int y5_days[y5_n];                     
-  real y5_se;
-
-  int y6_n;
-  real y6_values[y6_n];       
-  int y6_days[y6_n];                     
-  real y6_se;
+  // int y5_n;
+  // real y5_values[y5_n];       
+  // int y5_days[y5_n];                     
+  // real y5_se;
+  // 
+  // int y6_n;
+  // real y6_values[y6_n];       
+  // int y6_days[y6_n];                     
+  // real y6_se;
 
-  int y7_n;
-  real y7_values[y7_n];       
-  int y7_days[y7_n];                     
-  real y7_se;
+  // int y7_n;
+  // real y7_values[y7_n];       
+  // int y7_days[y7_n];                     
+  // real y7_se;
 
 }
 
 parameters {
   vector[election_days[number_elections]] mu;         // 
-  real d[7];                      // polling effects
+  real d[4];                      // polling effects
   real<lower=0> sigma;            // sd of innovations
 }
 
@@ -75,9 +75,8 @@ model {
   y2_values ~ normal(mu[y2_days]   + d[2],  y2_se * inflator);
   y3_values ~ normal(mu[y3_days]   + d[3],  y3_se * inflator);
   y4_values ~ normal(mu[y4_days]   + d[4],  y4_se * inflator);
-  y5_values ~ normal(mu[y5_days]   + d[5],  y5_se * inflator);
-  y6_values ~ normal(mu[y6_days]   + d[6],  y6_se * inflator);
-  y7_values ~ normal(mu[y7_days]   + d[7],  y7_se * inflator);
+  // y5_values ~ normal(mu[y5_days]   + d[5],  y5_se * inflator);
+  // y6_values ~ normal(mu[y6_days]   + d[6],  y6_se * inflator);
+  // y7_values ~ normal(mu[y7_days]   + d[7],  y7_se * inflator);
 
 }
-

model-2pp/model-interpretation.R

@@ -1,4 +1,6 @@
-load("output/model_2pp_2019051801014.rdata")
+load("output/model_2pp.rdata")
+
+the_caption <- ""
 
 #-------------------------House effects-------------------
 
@@ -76,33 +78,26 @@ sd(election_sims)
 # individual level swing statndard deviation is 3.2
 # last election was 47.52 ALP 
 
-alp_v_coal <- oz_pendulum_2019[1:142, ] %>%
+alp_v_coal <- oz_pendulum_2022%>%
+  filter(!division %in% c("Cooper", "Grayndler", "Indi", "Mayo", "Warringah", "Kennedy", "Melbourne", "Clark")) %>%
   mutate(margin = ifelse(party_against == "ALP", -margin, margin),
          pos_winner = "ALP",
          alt_winner = "Lib/Nat", 
          alp_v_coal = 1) %>%
-  select(state, division, pos_winner, margin, alt_winner, alp_v_coal) %>%
-  rbind(tibble(
-    state = "VIC",
-    division = "Indi",
-    pos_winner = "ALP",
-    margin = -8.8,
-    alt_winner = "Lib/Nat",
-    alp_v_coal = 1
-  ))
-
-alp_v_other <- oz_pendulum_2019 %>%
-  filter(division %in% c("Wills", "Cooper", "Grayndler", "Clark")) %>%
+  select(state, division, pos_winner, margin, alt_winner, alp_v_coal)
+
+alp_v_other <- oz_pendulum_2022 %>%
+  filter(division %in% c("Cooper", "Grayndler", "Clark")) %>%
   mutate(pos_winner = "ALP",
-         margin = c(4.9, 1.3, 15.8, -17.8),
-         alt_winner = c("Grn", "Grn", "Grn", "Ind"), 
+         margin = c(14.6, 16.3, 22.1),
+         alt_winner = c("Grn", "Grn", "Ind"), 
          alp_v_coal = 0) %>%
   select(state, division, pos_winner, margin, alt_winner, alp_v_coal)
 
-other_v_coal <- oz_pendulum_2019 %>%
-  filter(division %in% c("Wentworth", "Mayo", "Kennedy", "Melbourne")) %>%
+other_v_coal <- oz_pendulum_2022 %>%
+  filter(division %in% c("Indi", "Mayo", "Warringah", "Kennedy", "Melbourne")) %>%
   mutate(coal_margin = margin,
-         pos_winner = ifelse(incumbent %in% c("ALP", "KAP", "NXT"), 
+         pos_winner = ifelse(incumbent %in% c("ALP", "KAP", "CA"), 
                              incumbent, 
                              str_to_sentence(incumbent)),
          alt_winner = "Lib/Nat", 

pkg/DESCRIPTION

@@ -19,4 +19,4 @@ Suggests:
 License: GPL-3
 Encoding: UTF-8
 LazyData: true
-RoxygenNote: 6.1.1
+RoxygenNote: 7.1.2