R's na.rm equivalent in Julia for ignoring missing values during inference?

New to Julia
1

Do we have anything like r’s na.rm attribute in Julia for Turing.jl for ignoring missing values? I have a data with missing values which is causing issue during sampling.

2

Don’t know Turing.jl, but the julia way to “ignore” missing values is to “skip” them:

julia> x = [1,missing,3]
3-element Vector{Union{Missing, Int64}}:
 1
  missing
 3

julia> sum(x)
missing

julia> sum(skipmissing(x))
4
2 Likes
3

This I know but I’m looking for something like this as shown below.

 for i in 1:n
      for j in 1:size(data[i])[2]
         
        data[i][:,j] ~ MvNormal(predicted[i] ,  σ)
   
      end 
    end

Thanks for reply though :slight_smile:

4

How valid is this ?

for j in 1:size(data)[2]
         for k in 1:size(data)[1]

             if ismissing(data[k,j]) == true
                 continue
             end
          
             data[k,j] ~ Normal(predicted[k] ,  σ*(data[k,j] .+10^-5))

        end
end
5

Basically yes, but the more idiomatic way may be with a guard like this:

for j in axes(data, 2)     
    for k in axes(data, 1)
        ismissing(data[k, j]) && continue 
      
        data[k,j] ~ Normal(predicted[k] ,  σ*(data[k,j] .+10^-5))
    end
end