Skip to content

NEWS.md

Latest commit

 

History

History
196 lines (120 loc) · 5.66 KB

NEWS.md

File metadata and controls

196 lines (120 loc) · 5.66 KB

refund 0.1-40

Breaking changes

  • The default sandwich option for pffr() is now sandwich = "cluster" (previously "none"). Existing code calling pffr() without explicit sandwich= will now get cluster-robust standard errors and confidence intervals. Set sandwich = "none" to restore the previous behavior. A one-time informational message() is printed when sandwich is not explicitly supplied.
  • pffrGLS() / pffr_gls() are deprecated and now error. Their GLS-based covariance correction produced poorly calibrated inference. Use pffr() with sandwich = "cluster" (default) or sandwich = "cl2" instead.

Function renames (old names deprecated)

  • pffrSim() → pffr_simulate() (old name warns via .Deprecated())
  • coefboot.pffr() → pffr_coefboot()
  • qq.pffr() → pffr_qq()
  • pffr.check() → pffr_check()

New features

  • pffr() modularization: internal refactor into prepare → fit → postprocess pipeline (pffr_prepare(), pffr_build_label_map(), etc.) for better maintainability.
  • pffr_simulate() (formerly pffrSim()) now supports a formula-based interface for specifying simulation models (e.g., pffr_simulate(Y ~ ff(X1) + xlin, effects = list(X1 = "cosine"))). The new interface provides:
    • Customizable effect functions via preset libraries or user-defined functions
    • Access to true coefficient functions via the truth attribute
    • Support for non-Gaussian responses via the family argument
  • The scenario argument in pffr_simulate() is deprecated. Use the formula interface instead. Legacy code using scenario will continue to work but will emit a deprecation warning.
  • pffr(..., sandwich = "cl2") and coef.pffr(..., sandwich = "cl2") now support leverage-adjusted cluster-robust covariance (Bell-McCaffrey style CL2), including for family = mgcv::gaulss().
  • coef.pffr() now supports confidence intervals via ci = "pointwise" or ci = "simultaneous" in addition to standard errors. Simultaneous intervals are computed with a coefficient-level Gaussian simulation and max-|t| calibration over each smooth term's evaluation grid.
  • AR(1) support improvements: pffr() now automatically switches to algorithm = "bam" and method = "fREML" when rho is supplied, and sets discrete = TRUE for non-Gaussian families.
  • Removed dependency on mgcv::plot.random.effect

refund 0.1-37

  • Fixed invalid email address for Yakuan Chen

refund 0.1-36

  • Fix threshold for small sigma2 in mfpca.face function to ensure accurate score estimation for level1 scores
  • Added parameters npc2 and pve2 to mfpca.face to allow for user to specify separate npc or pve for level 2 decomposition

refund 0.1-35

  • One line fix to pfr that allows pfr to be called from within another function
  • Pull request on URL updates in documentation

refund 0.1-34

  • Added pve to what is returned by mfpca.face
  • Change names in list of what is returned by mfpca.face to be the same as mfpca.sc
  • Updated email address for Erjia Cui
  • Changed Erjia from contributor to author

refund 0.1-33

  • Added pve to what is returned by fpca.sc and fpca.face

refund 0.1-32

  • Simon Wood removed "pers" argument from mgcv::plot.random.effect. Removed reference to pers from plot_pfr.gam to avoid documentation and code being inconsistent.

refund 0.1-31

  • added COVID19 datasets
  • minor bug fix for mfpca.face
  • updated email address for package maintainer

refund 0.1-30

  • removed import lattice::qq from pffr-methods.R

refund 0.1-28

  • added periodic spline option for fpca.face
  • changed if(class(object) != "string") to if(!inherits(object, "string")) in ccb.fpc.R and fosr.perm.test.R files to fix Note.

refund 0.1-27

  • bug fix for pfr models without intercept (thx, @ZheyuanLi)

refund 0.1-26

  • New function, mfpca.face, which is a faster version of mfpca.sc

refund 0.1-25

  • Minor bug fixes in fpca.face to patch error when npc = 1

refund 0.1-24

  • Minor bug fixes in pfr to patch error in R version 4.1
  • Updated documentation URLs

refund 0.1-23

  • Minor bug fixes in fgam examples for upcoming R release

refund 0.1-22

  • Fixes bugs
    • Commented out option useSymm = TRUE in tests for fpca.sc

refund 0.1-21

  • Fixes a bug in fpc() due to release of R 4.0.0 that changes the following:
 R> class(matrix(1 : 4, 2, 2))
 [1] "matrix" "array" 

(and no longer just "matrix" as before), and that conditions of length
greater than one in 'if' and 'while' statements executing in the package
being checked give an error.
  • change email address for maintainer

refund 0.1-20

  • fix minor bug in rlrt.pfr.R

  • change maintainer from Rayman Huang to Julia Wrobel

refund 0.1-19

  • updates for compatibility with mgcv 1.8-23 (#69 etc.)

  • fixed fpcr for scalar covariates (#76)

  • now re-exports cmdscale_lanczos

refund 0.1-15

  • homogenized inputs/outputs to most fpca.XXX functions

  • fix documentation for fpca.face

  • added pco ridge regression see ?poridge

refund 0.1-14

  • add fpca.lfda() function

  • add functions and data set from refund.shiny package

refund 0.1-13

  • add mfpca.sc() function.

  • add example for DTI data.

  • export pfr_old().

  • fix documentation and add warning message for rlrt.pfr().

refund 0.1-12

  • new pfr() function. The new pfr() function merged the old pfr and fgam functionality.

  • Switch to roxygen2 documentation

  • allow an "argvals" argument to all relevant functions for consistent strucuture throughout the package

  • refactor input and output argument for fosr().

  • add vignettes.

  • bug fix in fpca.face().

  • export expand.call().

  • add cross-sectional FoSR using GLS, variational Bayes and Gibbs sampler