Defining specific eta's

1

Instead of writing with MvNormal, I am using Normal in the @random block. Please check is this code is correct.

    @param begin
      tvcl     ∈ RealDomain(lower=0)
      tvvc     ∈ RealDomain(lower=0)
      Ω        ∈ PDiagDomain(2)
      σ_add    ∈ RealDomain(lower=0.001)
    end
  
    @random begin
      ηCL        ~ Normal(0, Ω)
      ηVc        ~ Normal(0, Ω)

     end
2

Welcome to the discussion board @jayashreed1412 !

while the specification of the @random block is correct, you need to adjust the @param block. Let me show the two ways of writing the same model.

Using PDDiagDomain for your variances.

In this case, the η's are sampled from a MvNormal distribution of Ω which is defined as Positive Definite Diagonal (PDDiag) Matrix Domain. The number inside the parenthesis of PDDiagDomain represents the number of diagonal elements (the number η's ) in the block.
Important Note: The Ω here represents a variance for the MvNormal.

    @param begin
      tvcl     ∈ RealDomain(lower=0)
      tvvc     ∈ RealDomain(lower=0)
      Ω        ∈ PDiagDomain(2)
      σ_add    ∈ RealDomain(lower=0.001)
    end
  
    @random begin
      η ~ MvNormal(Ω)
     end

When you write the model this way, the η's are a vector and hence have to be indexed to access. So, your @pre block would like this where η[1] and η[2] below represent the samples from diagonals of the matrix.

    @pre begin
       CL = tvcl * exp(η[1])
       Vc = tvvc * exp(η[2])
    end

Using RealDomain for your variances

In this case, the η’ for each parameter is sampled from a Normal distribution of ω which is defined as Real Domain which works like any other parameter. Each η will have its corresponding ω.
Important Note: The ω here is the standard deviation of the Normal distribution.

    @param begin
      tvcl     ∈ RealDomain(lower=0)
      tvvc     ∈ RealDomain(lower=0)
      ωCL      ∈ RealDomain(lower=0)
      ωVc      ∈ RealDomain(lower=0)
      σ_add    ∈ RealDomain(lower=0.001)
    end
  
    @random begin
      ηCL        ~ Normal(0.0, Ω)
      ηVc        ~ Normal(0.0, Ω)
     end

When you write the model this way, the η can be named (e.g. ηCL) which is easier to read and remember than an indexed version. So, your @pre block would like this where ηCL and ηVc below represent the samples from a Normal distribution with mean 0.0 and variance ω (Note mean should be 0.0 and not 0)

    @pre begin
       CL = tvcl * exp(η[1])
       Vc = tvvc * exp(η[2])
    end
3

@vijay It is working thank you.