Materials for workshop, "Mixed-effects Models with Julia" at the 2017 New England Statistics Symposium, Friday, April 21, 2017

• It will help if participants install Julia (version 0.5.1), before the workshop. An alternative is to use the online service at https://juliabox.com, as described below.

• Julia can be installed from https://julialang.org/downloads/ or as the no-cost version of JuliaPro from https://juliacomputing.com.

• JuliaPro is a "batteries included" version of Julia with several of the more widely-used packages. Those using Windows or Mac OSX may find it more convenient than downloading the base release and adding packages.

• We will use Jupyter notebooks throughout the course.
• The interface to Jupyter notebooks is included with JuliaPro.
• Those who install Julia from the download site will need to add the IJulia package. See below for instructions on adding Julia packages.

• Julia packages are added with, e.g.

Pkg.add("MixedModels")

• It is a good practice to run

Pkg.update()

weekly or more frequently to ensure you have the latest package versions.

• To check which version of a package is installed, use, e.g.

Pkg.status("MixedModels")

• To see all of the directly and indirectly installed packages, use

versioninfo(true)

• In addition to MixedModels, you should install DataFramesMeta, FreqTables, Gadfly, IJulia, Rcall, and RData. Other packages we will use, (e.g. DataFrames, GLM) will be installed as requirements of these packages. Ensure that you have v0.8.0 or later of the MixedModels package.

• Julia packages are git repositories. At present registered packages must be available on https://github.com

• Registered packages are listed on https://pkg.julialang.org

• Documentation for packages is usually linked in the github repository. Search for the package name on https://pkg.julialang.org then click on the name to get to the github repository.

• In the package listing you will see that several packages are owned by github groups, such as JuliaStats, JuliaIO, JuliaOpt, JuliaDiffEQ, ... These packages can be expected to be well-maintained.

The notebooks (files ending in .ipynb) whose names start with numbers are intended to be read in the sequence.

• juliabox.com provides a browser interface to Julia through Jupyter notebooks.
  • login using one of the methods shown there
  • the Sync tab allows access to github.com repositories or Google Drive directories.
  • the Files tab provides uploading/downloading of files or folders
  • the Console provides shell access
  • Jupyter notebooks can be run from the Jupyter tab
  • the first couple of sections of the tutorial may be of interest. Later sections are Python-oriented and perhaps of less interest.

• the RData package can used to read saved .RData or .rda files. It does not require a local installation of R

• if you have a local installation of R, the RCall package allows you to run and communicate with an R process from within julia.

• feather is a language-agnostic form of dataframe storage. The R package is feather, Julia package is Feather and Python package is feather-format.

• when reading a Feather file in Julia use nullable=false for the time being

• there is the readtable function in DataFrames and a CSV package for reading CSV files.

• the PyCall package essentially integrates Python into Julia.
• the Pandas package provides access to the pandas package for Python