Question

        install.packages("MASS")
library("MASS")
attach(bacteria)
nrow(bacteria)
install.packages('BAS')
library(BAS)
set.seed(12345)

glimpse(bacteria)

help(bacteria)
cor = glm(y~ap+hilo+trt+week, family= "binomial", data = bacteria)
summary(cor)

plot(cor)
resid(cor)
pairs(bacteria)
plot(y~week, data = bacteria)

beta = coef(cor)
resid = residuals(cor)
n = length(resid)
mse = 1/(n-2)*sum((resid^2))
mse
output = summary(cor)$coef[,1:4]
output #OLS estimates
par(mfrow = c(2,2))

corred = glm(y~week, family = "binomial", data = bacteria)#Using only statistically significant variables from cor. 
summary(red)

# A function to evaluate the log of the posterior density
logP=function(y,X,b,b0,varB){
  Xb=X%*%b
  theta=exp(Xb)/(1+exp(Xb))
  logLik=sum( dbinom(x=y,p=theta,size=1,log=T)  )
  logPrior=sum(  dnorm(x=b,sd=sqrt(varB),mean=b0,log=T))
  return(logLik+logPrior)
}

logisticRegressionBayes=function(y,X,nIter=100000,V=.02,varB=rep(10000,ncol(X)),b0=rep(0,ncol(X))){

  ####### Arguments #######################
  # y  a vector with 0/1 values
  # X  incidence matrix of effects
  # b0,varB, the prior mean and prior variance bj~N(b0[j],varB[j])
  # V the variance of the normal distribution used to generate candidates~N(b[i-1],V)
  # nIter: number of iterations of the sampler
  # Details: generates samples from the posterior distribution of a logistic regression using a Metropolis algorithm
  #########################################

  # A matrix to store samples
  p=ncol(X)
  B=matrix(nrow=nIter,ncol=p)
  colnames(B)=colnames(X)

  # A vector to trace acceptance
  accept=matrix(nrow=nIter,ncol=p,NA)
  accept[1,]=TRUE 

  # Initialize
  B[1,]=0
  B[1,1]=log(mean(y)/(1-mean(y)))
  b=B[1,]
  for(i in 2:nIter){

    for(j in 1:p){
      candidate=b
      candidate[j]=rnorm(mean=b[j],sd=sqrt(V),n=1)

      logP_current=logP(y,X,b0=b0,varB=varB,b=b)
      logP_candidate=logP(y,X,b0=b0,varB=varB,b=candidate)
      r=min(1,exp(logP_candidate-logP_current))
      delta=rbinom(n=1,size=1,p=r)

      accept[i,j]=delta

      if(delta==1){ b[j]=candidate[j] }
    }
    B[i,]=b
    if(i%%1000==0){
      message(" Iteration ",i)
    }

  }

  return(list(B=B,accept=accept))


}


Z=as.matrix(model.matrix(y~ap+hilo+trt+week, data=bacteria))
samples=logisticRegressionBayes(bacteria$y,X=model.matrix(y~ap+hilo+trt+week, data=bacteria),nIter=55000)

我正在尝试查找数据集的后验和先验。我正在使用细菌数据集；

This data frame has 220 rows and the following columns:

y
presence or absence: a factor with levels n and y.

ap
active/placebo: a factor with levels a and p.

hilo
hi/low compliance: a factor with levels hi amd lo.

week
numeric: week of test.

ID
subject ID: a factor.

trt
a factor with levels placebo, drug and drug+, a re-coding of ap and hilo.

当我尝试运行最后一行时，出现错误：

Error in dbinom(x = y, p = theta, size = 1, log = T) : 
  Non-numeric argument to mathematical function
In addition: Warning messages:
1: In mean.default(y) : argument is not numeric or logical: returning NA
2: In mean.default(y) : argument is not numeric or logical: returning NA
3: In rnorm(mean = b[j], sd = sqrt(V), n = 1) :
 Show Traceback

 Rerun with Debug
 Error in dbinom(x = y, p = theta, size = 1, log = T) : 
  Non-numeric argument to mathematical function

以下是我的数据：各种因素，一些整数。

Observations: 220
Variables: 6
$ y    <fct> y, y, y, y, y, y, n, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, y, n, n, n, y, n, y, y, y, y, y, y, y, ...
$ ap   <fct> p, p, p, p, a, a, a, a, a, a, a, a, a, p, p, p, p, p, p, p, p, p, p, a, a, a, a, a, a, a, a, a, p, p, p, p, p, p, p, p, ...
$ hilo <fct> hi, hi, hi, hi, hi, hi, hi, hi, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, lo, hi, hi, hi, ...
$ week <int> 0, 2, 4, 11, 0, 2, 6, 11, 0, 2, 4, 6, 11, 0, 2, 4, 6, 11, 0, 2, 4, 6, 11, 0, 2, 4, 11, 0, 2, 4, 6, 11, 0, 2, 4, 6, 11, 0...
$ ID   <fct> X01, X01, X01, X01, X02, X02, X02, X02, X03, X03, X03, X03, X03, X04, X04, X04, X04, X04, X05, X05, X05, X05, X05, X06, ...
$ trt  <fct> placebo, placebo, placebo, placebo, drug+, drug+, drug+, drug+, drug, drug, drug, drug, drug, placebo, placebo, placebo,...

我正在尝试查找此数据集的先验和后验。当我使用给定功能时，它将对我不起作用。我提供了错误消息，原始代码以及数据组成。我希望你能帮助我。

谢谢。

Answer 1

代码为预期为数字（dbinom）的参数传递了y一个因数（x = y）。尝试将y转换为数字，例如

y_num <- as.numeric(y == 'y')

并使用它

dbinom(x=y_num, p=theta, size=1, log=T)

关于R中传递给数学函数的非数字参数的错误

1 个答案: