Question

我正在尝试使用brms为我的传销设置先验。我已经使用先前设置的模型运行了我的模型，没有出现错误消息，现在想使用pp_check对其进行检查。我收到一个'sdata [[paste0（“ Y”，usc（resp））]]中的错误：下标超出范围”错误，找不到关于为什么发生这种情况的任何提示。谢谢！

编辑：我已经检查了fit的结构，并且只有init变量为0，我认为应该是这种情况，因为我将初始化参数设置为0？否则，我看不到任何问题。

自从谷歌搜索什么都没做以来，我什么也不能尝试。

library(brms)
df <- data.frame( subjno = as.factor(c('sub-01', 'sub-01','sub-01','sub-01','sub02','sub02','sub02','sub02')),
              L1 = c(0.898922096, -0.673393065, -2.240150247,-0.932520537, -0.472701111, -0.188825324,0.808675919, 0.293666248), 
              L2 = c(0.64888, 2.0891, -0.655322708, 0.007098555, -0.648887797, -0.249716343, -0.698128026,0.119511014), 
              W1 = c(0.5,0.5,-0.5,-0.5,0.5,-0.5,0.5,-0.5), W2 = c(0.5,-0.5,0.5,-0.5,0.5,0.5,-0.5,-0.5),
              t = as.factor(c(12,23,34,45,12,23,34,45)))

ff_s = brmsformula(cbind(L1,L2) ~ W1 * W2 * t +  
                     (W1*W2* t|p|subjno)) 

get_prior(formula = ff_s, family = gaussian(),
      data = df)

pp_s <- c(set_prior('normal(0,1)', class = "b"),  
    set_prior("normal(0,10)", class = "sd", resp = 'L1'), 
    set_prior("normal(0,10)", class = "sd", resp = 'L2'), 
    set_prior("normal(0,5)", class = "sigma",resp = 'L1'), 
    set_prior("normal(0,5)", class = "sigma",resp = 'L2'),
    set_prior("normal(0,10)", class = "Intercept", resp = 'L1'),
    set_prior("normal(0,10)", class = "Intercept", resp = 'L2'),
    set_prior("lkj(3)", class = "cor"))

fit_s <- brm(formula = ff_s, 
          data = df, family = gaussian(),
          prior = pp_s,
          chains = 6, cores = 3,
          iter = 2e3, warmup = 1e3,
          init = 0,
          sample_prior = "only")

pp_check(fit_s)

Answer 1

我发现我在调用pp_check函数时未指定我感兴趣的级别，问题已解决！

如何使用brms软件包修复R中的“ sdata [[paste0（“ Y”，usc（resp））]]错误：下标超出范围”

1 个答案: