gretaパッケージを活用したベイズモデリングの紹介

model1

# library
library(greta)
library (bayesplot)
library(tidyverse)

# data
y <- as_data(iris$Sepal.Length)
Species <-as.numeric(iris$Species)

# variables and priors
mu <- normal(0, 10, dim = 3)
sd <- uniform(0, 10, dim = 3)

# operations
mean <- mu[Species]
sd_species <- sd[Species]

# likelihood
distribution(y) <- normal(mean, sd_species)

# defining the model
m <- model(mu, sd)

# plotting
plot(m)

# sampling
draws <- mcmc(m, n_samples = 1000)
summary(draws)


#result plot
mcmc_trace(draws)
mcmc_intervals(draws)

model2

# library
library(greta)
library (bayesplot)
library(tidyverse)

# data
x <- as_data(iris$Petal.Length)
y <- as_data(iris$Sepal.Length)
Species <-as.numeric(iris$Species)
n_species <- length(unique(Species))

# variables and priors
mu <- normal(0, 10, dim = n_species)
sd <- uniform(0, 10, dim = n_species)
coef <- normal(0, 10, dim = n_species)

# operations
mean <- mu[Species] + coef[Species] * x
sd_species <- sd[Species]

# likelihood
distribution(y) <- normal(mean, sd_species)

# defining the model
m <- model(mu, sd, coef)

# plotting
plot(m)

# sampling
draws <- mcmc(m, n_samples = 1000)
summary(draws)


#result plot
mcmc_trace(draws)
mcmc_intervals(draws)