Feature list for 0.1.0 release

[storopoli]

• [x] turing_model function:

  turing_model(formula, data;
      family=Normal,
      priors=DefaultPriors(),
      standardize=false)

• [x] Close interface with DataFrames.jl and CategoricalArrays.jl using the Tables.jl API:
  • [x] Reading the length(unique(x)) of a group-level intercept/slope
  • [x] Auto dummy variable ($K - 1$) with categorical vectors by reading the levels and using the first level as baseline.
• [x] Likelihoods:
  • [x] Linear Regression: Gaussian and TDist (IdentityLink)
  • [x] Logistic Regression: Bernoulli
  • [x] Count Data: Poisson and NegativeBinomial (LogLink)
• [x] Priors:
  • [x] Default Priors and Custom Priors
• [x] Group-level effects (Hierarchical Models):
  • [x] Non-Centered Parameterization and recompute the parameter back in the return statement
  • [x] Varying-Intercept: (1 | group)
• [x] Case Studies showcasing:
  • [x] @formula syntax
  • [x] random effects: random-intercept
  • [x] likelihoods
  • [x] CategoricalArrays.jl and DataFrames.jl (Tables.jl) interfaces
  • [x] Use the {rstanarm} datasets (check LICENSE) and the sleep dataset from {brms}

[phipsgabler]

One of these models embedded as a @submodel in another Turing model would be a cool show case and interaction test, I think.

[storopoli]

Thanks, do you have good resources on @submodel. I want o learn more about it, never used.

[phipsgabler]

They are unfortunately not at all documented, but the PR should cover the usage.