glmmstan: Generalized liner mixed model in Rstan

> library(glmmstan)

Loading required package: rstan
rstan (Version 2.8.2, packaged: 2015-11-26 15:27:02 UTC, GitRev: 05c3d0058b6a)
For execution on a local, multicore CPU with excess RAM we recommend calling
rstan_options(auto_write = TRUE)
options(mc.cores = parallel::detectCores())

Attaching package: 'rstan'

The following object is masked from 'package:raster':

    extract

The following object is masked from 'package:magrittr':

    extract

> data("baseball")

バージョン: 1.31

関数名 概略
baseball Baseball player's perfomance data
glmmstan Generalized liner mixed model in Rstan using glmer-style formulas
glmmstan-package Generalized liner mixed model in Rstan
iccstan intra-class correlations in rstan

baseball

野球選手の成績や所属チームのデータセット

> data("baseball")
> head(baseball)

  player team league   AVR ATbase ATbats HIT HR RBI steal walk   K salary
1      1    1      1 0.324    685    596 193 29  89    15   74  95     80
2      2    1      1 0.305    637    580 177 10  69     2   43  62     85
3      3    1      1 0.316    597    547 173 23  90    10   48 103     60
4      4    1      1 0.310    464    422 131 17  79     2   36  54     64
5      5    1      1 0.301    446    366 110 31  69     2   75  95    297
6      6    1      1 0.298    365    325  97  5  41     0   26  51     40
  salary_log Cluster
1   4.382027       1
2   4.442651       1
3   4.094345       2
4   4.158883       2
5   5.693732       3
6   3.688879       2

glmmstan

RStanを利用した一般化線形混合モデルの実行

Arguments

  • formula_str
  • data
  • family
  • center
  • offset
  • codeonly
  • dataonly
  • modelonly
  • cauchy
  • parallel
  • cores
  • iter
  • warmup
  • thin
> fit <- glmmstan(salary_log ~ 1, data = baseball, family = "gaussian")
> output_result(fit)
> 
> fit <- glmmstan(HR ~ 1 + (1 | team), data = baseball, family = "poisson")
> output_result(fit)
> 
> fit <- glmmstan(salary ~ HR + K + (HR | team), data = baseball, family = "lognormal")
> output_stan(fit)
> 
> # 打率がリーグごとに異なるか
> df_team <- baseball %>% dplyr::filter(team %in% c(1, 2))
> fit <- df_team %>% glmmstan(cbind(HIT, ATbats - HIT) ~ team + (1 | player),
+                              data = .,
+                              family = "binomial")

> baseball %>% ggplot(aes(factor(team), AVR)) + geom_boxplot()
> baseball %>% ggplot(aes(factor(team), HR)) + geom_boxplot()

参考