3

After estimating the conditional probability in Bayesian networks, I asked the probability of one node ("Inlet_gas_total_pressure") as follows;

bn.mle.before$"Inlet_gas_total_pressure"

Parameters of node Inlet_gas_total_pressure (multinomial distribution)

Conditional probability table:

      no      yes 
0.843127 0.156873 

bn.mle.before$"Inlet_gas_total_pressure"$prob

      no      yes 
0.843127 0.156873

I want to change the probability value of "yes" from 0.156873 to 0.4.
How can I do that ? The following was my trial, but fails.

bn.mle.before$"Inlet_gas_total_pressure" <- list(prob=c("no"=0.6, "yes"=0.4))

Error in check.fit.dnode.spec(value, node = name) : the conditional probability distribution of node Inlet_gas_total_pressure must be a table, a matrix or a multidimensional array.

Please help me.

Sotos
  • 51,121
  • 6
  • 32
  • 66
Hong-Bae Jun
  • 31
  • 1
  • 1
    The error suggests it should not be a list. Have a look at `str(bn.mle.before$"Inlet_gas_total_pressure")`, to see what sort of object it is , or even `dput(bn.mle.before$"Inlet_gas_total_pressure")` , and then construct your new table the same way – user20650 Sep 21 '16 at 10:36

2 Answers2

1

I experienced the same problem. Here is some toy example that will show you how to save the day.

library(bnlearn)
Learning.set4=cbind(c("Yes","Yes","Yes","No","No","No"),c("Blue","Green","Blue","Green","Green","Green"),c(9,10,8,3,2,1))
Learning.set4=as.data.frame(Learning.set4)
Learning.set4[,c(3)]=as.numeric(as.character(Learning.set4[,c(3)]))
colnames(Learning.set4)=c("Cause1","Cause2","Cons")
b.network=empty.graph(colnames(Learning.set4))
struct.mat=matrix(0,3,3)
colnames(struct.mat)=colnames(Learning.set4)
rownames(struct.mat)=colnames(struct.mat)
struct.mat[2,3]=1
struct.mat[1,3]=1
bnlearn::amat(b.network)=struct.mat
haha=bn.fit(b.network,Learning.set4)

print(haha$Cause1$prob)

T=haha$Cause1$prob
T[[1]]=0.8
T[[2]]=0.2

haha$Cause1=T
print(haha$Cause1$prob)

I succesfully changed the probabilities for the node Cause1

Cheers

Joel H
  • 173
  • 15
0
# a similar example

fit=bn.fit(dag,traindata)

# Below I want to set any zero prob to something small

for (i in 1:10) {
  my=fit[[i]]
  idx=which(my$prob==0)
  if (length(idx)>0){
    for (j in idx ) {
      my$prob[[j]]=0.001
      my$prob[[j-1]]=1-0.001
    }
  }
  fit[i]=list(my)
}
joelc
  • 1