不倫報道間隔を予想する

furin_span2016.R

## 2016年度の不倫報道データ
dataset <- as.data.frame(
matrix(c("ベッキー","2016-01-07",
         "狩野英孝","2016-02-03",
         "宮崎謙介議員","2016-02-10",
         "桂文枝","2016-03-16",
         "石井竜也","2016-03-17",
         "乙武洋匡","2016-03-23",
         "とにかく明るい安村","2016-04-01",
         "NMB48木下春奈","2016-05-08",
         "ファンキー加藤","2016-06-07",
         "三遊亭円楽","2016-06-10",
         "荻上チキ","2016-07-06",
         "小倉優子の夫","2016-08-02",
         "中村橋之助","2016-09-15",
         "ジュビロ磐田の藤田義明","2016-10-04",
         "浦沢直樹","2016-10-10",
         "矢島悠子アナウンサー","2016-10-13",
         "ラモス瑠偉","2016-10-18",
         "小渕健太郎","2016-10-20"
),ncol=2,byrow=T))
names(dataset) <- c("who","date")

# POSIXct型に変換
furin <- as.POSIXct(dataset$date)
# 不倫報道が生じるまでの日数を計算
span <- c()
for(i in 1:(length(furin)-1)){
  span[i] <- difftime(furin[i+1],furin[i])
}


# Stanで予測する
library(rstan)
options(mc.cores = parallel::detectCores())

# 「報道が生じる」のがポアソン分布，ポアソン分布で
#  次の観測が生じるまでの感覚は指数分布に従うと考えられる

stancode <- '
data{
  int<lower=0> N;
  real<lower=0> X[N];
}
parameters{
  real<lower=0> lambda;
}
model{
  target += exponential_lpdf(X|lambda);
}
generated quantities{
  real<lower=0> mu;
  mu = inv(lambda);
}
'

# 推定
datastan <- list(N=length(span),X=span)
model <- stan_model(model_code=stancode)
fit <- sampling(model,data=datastan)

print(fit)