我正在使用列车和验证变量之间的GLM进行回归建模 . 所有变量都是数值 . 但是当我尝试与错误运行相关时出现错误：'y'必须是数字..代码如下：

data <- read.csv(file="fielddatabase.csv", sep=",", header=T)
attach(data)
summary(data)

###############################
## Prepare Bootstrap samples
###############################

set.seed(550)

# create empty lists in which subsets can be stored
train <- list()
validation <- list()

# set the bootstrap parameters
N = length(data[,9]) # N° of observations
B = 500             # N° of bootstrap iterations

# start loop
for(i in 1:B){

  # create random numbers with replacement to select samples from each group
  idx = sample(1:N, N, replace=TRUE)

  # select subsets of the five groups based on the random numbers
  train[[i]] <- data[idx,]
  validation[[i]] <- data[-idx,]
}

######################################
## start regression modelling with GLM
######################################
# create empty lists in which the models accuracies can be stored
# Total
ID.nb<-list()
Obs.nb<-list()
Pred.nb<-list()
r2.nb<-list()
rmse.nb<-list()
Nrmse.nb<-list()
imp.nb<-list()
bias.nb <- list()


# Run GLM
for(i in 1:B)
  TRAIN <- train[[i]] 
  VALIDATION <- validation[[i]]
  len<-length( VALIDATION[,1])

  # store and select the observations
  ID<-VALIDATION$ID
  ID.nb [[i]]<-ID
  obs <- VALIDATION[,2]
  Obs.nb[[i]]<-obs

  # run the GLM using Negative Binomial family. 

  GLM_total <- glm(Biomass ~  hmax + Nmax + hsd + hskewness + hkurtosis, data=TRAIN,  family=negative.binomial(theta=1 , link="log"))

  # predict Biomass values 
  Pred<-stats:::predict(GLM_total, newdata=VALIDATION, type="response")

  # store the model accuracies
  Pred.nb[[i]]<-Pred
  r2.nb[[i]]<-(cor(Pred, obs, method="pearson"))^2

这里的错误显示：

Error in cor(Pred, obs, method = "pearson") : 'y' must be numeric
r regression