Skip to content

Latest commit

 

History

History
53 lines (41 loc) · 2.23 KB

NEWS.md

File metadata and controls

53 lines (41 loc) · 2.23 KB

2016-01-27

  • Handle constant_item = FALSE correctly in wrangle()

2016-01-24

  • Bugfix: group order could be scrambled in ZZ and XX

2016-01-18

  • Bugfix: adjusted trial and success counts (n_vec, s_vec), after calculation, could be associated with the wrong group names
  • Switch to numeric time variable in dgirt output

2016-01-10

  • Handle item variables (defensively) as follows:
    • A numeric item variable with two unique values or an ordered factor item variable with two observed levels represents a binary choice in which the higher value or level is "success" and the lower value or level "failure"
    • Generally, numeric and ordered-factor item variables represent ascending ordinal choices
    • Other classes of item variable (e.g. factor, character) result in an error
    • Print messages explaining the handling of each item variable
  • Remove dependency on mcgv in plot_means()

2016-01-06

  • Bugfix: mean item outcomes could be calculated incorrectly, inflating success counts toward trial counts.

2015-12-30

Version bump to 0.0.10.

  • Functionality:
    • Apply more descriptive names to dgirt() results using the variable names originally passed to wrangle() and the levels of factors.
    • poststratify() is safer and more flexible. It takes new arguments strata, groups, and check_proportions; see the documentation.
    • Specify the algorithm for CmdStan to use. dgirt() passes new argument optimize_algorithm to CmdStan if method = "optimize"; one of "lbfgs" (the default), "bfgs" and "newton".
  • Documentation:
    • Switch to a README.Rmd that includes the "Getting Started" vignette content and drop the vignette.
  • Functions renamed:
    • run_dgirt() -> dgirt()
    • format_dgirt() -> wrangle()
  • New and renamed datasets:
    • rstan_output: example of dgirt() output for method = "rstan"
    • optimize_output: example of dgirt() output for method = "optimize"
    • states -> state_opinion
    • state_targets and targets -> state_demographics
  • Under the hood:
    • Switch to assertthat package from ad-hoc stop() calls
    • Speed up wrangle()
    • Bugfixes