Question about Bayesian Survival Analysis with Cox (proportional hazards) regression method in Julia with Turing

[EvoArt]

@RyanKuo

@model function cox_regression(death,exposure,x)
    λ0 ~Gamma(0.01, 0.01)
   # σ  ~Uniform(0.0, 10.0)
   # τ  =  σ^2
   # μβ  ~ Normal(0.0, 100)
   # β  ~ Normal(μβ,τ)
     β  ~ Normal(0.0,100)
    λ = exp.(β .* x).* fill(λ0,n_intervals)'
    μ  = exposure .*λ    
    for i in 1:length(death)
      death[i] ~ Poisson(μ[i])
    end
end

chain = sample(cox_regression(death,exposure,HSAUR[!,:CMetastized]), NUTS(0.65),1000)

Hi Ryan

Is this the exact code you tried to run?

I get an error

ERROR: ArgumentError: column name :CMetastized not found in the data frame; existing most similar names are: :Metastized

and then after correcting that

MethodError: no method matching *(::Float64, ::CategoricalArrays.CategoricalValue{String, UInt8})

These are all quick little fixes, but it would be much easier if you have a working MWE. one quick thing I noticed is that you use global variables in the model like n_intervals. In Julia you will tend to get much better performance if you pass those variables into the actual function.