1

我正在尝试使用“Liszt”包在贝叶斯框架中执行列表长度分析（基本上是多元逻辑回归）。这是一个只能从这里安装的 beta 包： http ://www.possinghamlab.org/images/LLA/Liszt_0.8-5_1.tar.gz

当我查看包时，rjags 的模型如下：

A3ModelFun <-
function(Species, Year, ListLength = rowSums(Species), sp.names = colnames(Species),
                           midYear = round(median(unique(Year))),
                           n.iter=50000, n.burnin=20000,
                           resultFolder = NULL) {

.folder <-
    function(tag = "ListLengthAnalysis") {
      paste(gsub("[^0-9]", "_", as.character(strptime(date(), "%a %b %d %H:%M:%S %Y"))), tag, sep="_")
    }

#checks
#Will not work if all the years aren't supplied
       if(length(Year) != nrow(Species)) { 
           stop("Number of years supplied not equal to number of lists") }              

#When there is only one species the analysis wont run ... 
      if(length(Species) == 1) {
        stop("Only one species") }

  #Assign default folder name if unspecified
  if(is.null(resultFolder)) resultFolder <- .folder()
  is.folder <- function(file) file.info(file)$isdir
  if(is.na(is.folder(resultFolder))) dir.create(resultFolder) else print("folder exists") 

## Set up a folder for the results     

#      if(file.exists(resultFolder))
#        if(!is.folder(resultFolder)) stop("File ", resultFolder, " exists and is not a folder") else
#          dir.create(resultFolder)

  #Assign default midYear if unspecified
  if(is.null(midYear)) midYear <- round(median(unique(Year)))

      ## Create a function to generate inits (outside the for loop as it is the same params for every model)
      inits <- function() list(a1=rnorm(1), a2=rnorm(1), a3=rnorm(1))

      ModelFile <- tempfile()
      LL <- ListLength
      no <- nrow(Species)                       ## need to set these
      my <- midYear
      nyear <- length(unique(Year))
      meanLogLL <- mean(log(ListLength))
      halfyr <- nyear/2


##Progress Report
      j <- 1
      tot <- length(sp.names)
      for (i in sp.names) {
        cat(i, "\n")
        cat(j, "out of", tot, "\n")
        j <- j+1
      }

#####List Length Analysis#####        
      for (i in sp.names) {  # changed this from for (i in sp.names[]) { & for (i in sp.names[i])

        ## Define data for each model (we can reuse this object for each model)
        LLAData.1sp <- list(pres=Species[, i], yr=Year, LL=ListLength)

        ## Define the JAGS model
        cat(sprintf('
                    model {               
                    for (i in 1:%s) {
                    pres[i] ~ dbern(p[i])
                    logit(p[i]) <- a1 +
                    a2 * (log(LL[i]) - %s) +
                    a3 * (yr[i] - %s)            
                    }

                    a1 ~ dnorm(0, 0.0001)
                    a2 ~ dnorm(0, 0.0001)
                    a3 ~ dnorm(0, 0.0001)


                    }', no, meanLogLL, my, my),
              file=ModelFile)

    ## Run a model and store it in an R object called model_spname (where spname changes with each model)
    assign(sprintf('model_%s', i),
           jags(LLAData.1sp, inits,
                parameters.to.save=c(
                  'a1', 'a2', 'a3'),
                n.iter=n.iter, n.burnin=n.burnin, ModelFile))

    ## Save the current model to a file with the same names
    save(list=sprintf('model_%s', i), file=file.path(resultFolder, sprintf('model_%s.RData', i)))

    ## Delete the current model to save space
    rm(list = sprintf('model_%s', i))
}

  file.remove(ModelFile)
  return(resultFolder)
  }

模型本身有很多不相关的东西，但我觉得最好放置整段代码。

当我用我的数据运行它时，我总是得到错误：

Error in jags.model(model.file, data = data, inits = init.values, n.chains = n.chains,  : 
  Error in node pres[5]
Node inconsistent with parents

我发现有更多人收到此错误消息，但它始终与与所用分布相关的数据有关（即伽马分布的负数据，二项式的泊松数据）。我有二项式分布式数据，但仍然得到错误。我不知道为什么。

我的模型如下（有关我的数据框的 20 行子样本，请参见下面的 dput，我仅在具有完整数据集的不同节点上得到相同的错误）：

library(dplyr)

df <- (dput)
sp.ll <-as.matrix(df %>% select(-Year, -Hok_uniek, -LL))

mod <- A3ModelFun(sp.ll, df$Year, df$LL)

数据（在驱动器上，否则我将超过最大字符数）：

https://drive.google.com/open?id=1etj_7IrEHJbYzBk9p_8tW6B-X0HpbLU9WMWMJL-QT_I

4

0 回答 0