Movatterモバイル変換

specr 1.0.0

CRAN release: [soon]

Breaking changes

• This new version introduces a completely new analytic frameworkwhich breaks with earlier versions of specr. Seethisvignette for a comprehensive tutorial on how to use this newframework.

• A new function calledsetup() is introduced andreplaces the originalsetup_specs() to make thespecification of analytical choices more intuitive and comprehensive.Most importantly, it allows to set up all specifications (now alsoincluding subset analyses, additions to the formula, etc.) a priori,i.e., before the estimation of all models.

  • Data, analytic choices, and even parameter extraction functions canbe specified before the fitting process (solving github issue #23 and#26).
  • Excludes some non-meaningful specifications automatically (e.g.,when control and independent or dependent variable are the same).
  • Allows to specify all combinations of control variables or to“simplify” and only include none, each individually and all together(github issue #11).
  • Resulting specification can be filtered using standardtidyverse functions such as e.g.,filter().This way, one can make sure a priori that only reasonable specificationsare actually included.
  • Produces an object of classspecr.setup, which can beinvestigated using e.g.,summary() orplot().

• The functionrun_specs() is replaced byspecr(), which now only wraps aroundsetup()to estimate all models. Most changes are related to increasing speed ofthe computations.

  • Most importantly, the estimation process can now be parallelized(based onfurrr) to reduce fitting time (finally solvinggithub issue #1).
  • Deals better and faster with subset analyses.
  • Produces an object of classspecr.object, which can beinvestigated using generic function such assummary() andplot().
  • Plotting function have been updated and e.g. also allow to groupspecification according to specific choices (github issue #19)

• For more information about these major changes and how to use thenew version of specr, see thisvignette.

• Please note that the functions from earlier versions are stillavailable, but deprecated. It is hence still possible to use the olderframework as implemented in version 0.2.1, but we suggest to move to thenew framework of version 0.3.0 due to its increased speed andflexibility.

Known issues/bugs

• So far none…

Acknowledgements

This version ofspecr was something I wanted to tacklefor a while. Although the previous versions seem to have embraced by thecommunity and several relevant papers used them in interesting ways,certain issues kept coming up (ranging from slow fitting process towanting to incorporate more complex analytical decisions and choices).This version is thus heavily inspired by the growing community ofscholars who provide feedback and suggestions viagithub. Thank you toall of you! But for this version, I want to thank two peopleparticularly:

• Big thanks go toMattiVuorre who explored ways to parallelize specification curve analysisin thisblogpost,which became the basis for the implementation in specr.

• Many thanks toKasperWelbers who contributed some essential code within the corefunction.

• Thanks also toJohannesGruber who provided support in the final stages of building thispackage.

specr 0.2.2

Development version released: 2020-12-04

Breaking changes

• None

Updates

• Some minor updates (related to gituhub issues) and bug fixes:

  • All combinations of control variables can be produced (by specifyingall.comb = TRUE; solved github issue #21)
  • run_specs() now allows to add sets of control variabels(github issue #11)
  • All plotting functions allow to choose which parameter to plot
  • More complete results based onbroom::tidy() andbroom::glance()

• The package further now allows to integrate:

  • random effects (see thisvignette)
  • latent measurement models (see thisvignette)

Known issues

• Still no parallelization (github issue #1)

specr 0.2.1

CRAN release: 2020-03-26

Updates

• First stable version

• Tested in several environments.

• Primary function isrun_specs(), which allows tospecify analytic choices and estimate all models acrossspecifications.

Known issues

• No parallelization of the fitting process (can take very long ifmodel fitting process is complex).

• Implementation of random effect modelling and structural equationmodeling potentially possible, but still unclear.

• Does not allow to specify all possible combinations of controlvariables (github issue #21).

• Does not allow to specify sets of control variables (github issue#11)