使用rstan运行回归估计

Question

我在R中通过rstan包使用stan。下面是我的模型。这个模型有一个交互术语，叫做X1 * X2

        library(rstan)
    library(bayesrules)
    data(weather_WU); head(weather_WU, 20)
    
    MCMC_Regression_Model =
        "
        data {
            int<lower = 0> n;
            vector[n] Y;
            vector[n] X1;
            vector[n] X2;
        }
        parameters {
            real beta0;
            real beta1;
            real beta2;
            real beta3;
            real<lower = 0> sigma;
        }
        model {
            Y     ~ normal(beta0 + beta1 * X1 + beta2 * X2 + 
                      beta3 * X1 .* X2, sigma);
            beta0 ~ normal(25, 5);
            beta1 ~ normal(0, 37.52);
            beta2 ~ normal(0, 0.82);
            beta3 ~ normal(0, 0.55);
            sigma ~ exponential(0.13);
        }
        "
    MCMC_Regression_SIMU =
        stan(model_code = MCMC_Regression_Model,
                data = list(n = nrow(weather_WU), 
                            Y = weather_WU[['temp9am']],    
                            X1 = weather_WU[['location']], 
                            X2 = weather_WU[['humidity9am']]
                        ),
                chains = 4,        
                iter = 5000 * 2,    
                seed = 84735
            )

这样，我就得到了下面的错误

    Error in mod$fit_ptr() : 
      Exception: variable does not exist; processing stage=data 
      initialization; variable name=X1; base type=vector_d  (in  
      'model1e9057045768_9ae288549657a6a89a994b0dc81a6d24' at 
        line 5)

这个错误表明variable does not exist，但是，变量在定义中几乎是存在的。

如果你能帮助我正确运行上面的代码，我将不胜感激。

Answer 1

我得到了基本上相同的错误(在重新安装rstan之后)-不同的散列，但相同的单词。所以我尝试在cmdstanr中直接做这件事。(虽然我不太使用Stan，但我最近的经验是，通过rstan的直接R-to-Stan接口最近变得非常不稳定。

## if necessary:
install.packages("cmdstanr",
        repos = c("https://mc-stan.org/r-packages/",
                                       getOption("repos")))
library(cmdstanr)
set_cmdstan_path("~/.cmdstan/cmdstan-2.27.0/") ## idiosyncratic/ if necessary

writeLines(MCMC_Regression_Model, con="tmpreg.stan")
mod <- cmdstan_model("tmpreg.stan")

fit <- mod$sample(
         data = list(n = nrow(weather_WU), 
                     Y = weather_WU[['temp9am']],    
                     X1 = weather_WU[['location']], 
                     X2 = weather_WU[['humidity9am']]
                     ),
         chains = 4,
         iter_warmup = 5000,
         iter_sampling = 5000,
         seed = 84735
         )

缩写会话信息：

R Under development (unstable) (2021-09-23 r80950)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Pop!_OS 20.10    

other attached packages:
[1] bayesrules_0.0.1     rstan_2.21.2         ggplot2_3.3.5       
[4] StanHeaders_2.21.0-7 cmdstanr_0.4.0