Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Commits on Apr 16, 2015
Commits on Apr 3, 2015
Commits on Dec 18, 2014
  1. To avoid warning when running R CMD CHECK (no visible global function

    definition for 'subject'), import the packages np, quantreg, and rgl
    in Imports: (DESCRIPTION) and NAMESPACE, remove require() calls in R
    code, add np.messages inclusion in zzz.R (will not be needed when/if
    npglpreg() migrates to the np package). Version number bumped up, no
    issues remaining with R CMD CHECK --as-cran. Demos all check out.
Commits on Dec 4, 2014
Commits on Aug 11, 2014
  1. Zhenghua's changes incorporated, updated version number and modded

    CHANGELOG, submitting to CRAN
Commits on Aug 9, 2014
  1. @ZhenghuaNie

    Fix bugs on setting seed.

    ZhenghuaNie committed
Commits on May 18, 2014
Commits on May 5, 2014
Commits on May 4, 2014
Commits on May 3, 2014
  1. Refining/modifying ridging in npglpreg... with changes to

    computationally more efficient varying coefficient implementation of
    the local polynomial estimator, ridiging needed to be modified,
    re-read Siefert and Gasser and taking convex combination of
    (potentially) shrunken local polynomial and local constant using the
    data-driven pointwise ridge factor
Commits on May 2, 2014
  1. Using solve() with default precision rather than call to

    chol2inv(chol()) for consistency (mixing can halt with error)
  2. Debugging npglpreg() with cv.aic... the move to one call and away from

    constant being the conditional mean to w'delta being the conditional
    mean required attention - believe issue is addressed, requires
Commits on Apr 26, 2014
  1. Testing for cv.shrink=TRUE and switching to FALSE if cv.aic selected

    to overcome existing glitch in shrinking with cv.aic
Commits on Apr 18, 2014
  1. Using qnorm(0.975) rather than 1.96 for accuracy, changed default

    cv.shrink to FALSE for npglpreg (may remain issues with shrinking for
Commits on Jan 22, 2014
  1. Incorporated resetting of DISPLAY_DEGREE=0 to frscvNOMAD and

    krscvNOMAD if not passed and if crs.messages=FALSE, updated version
    number and faq, ready to push to CRAN
Commits on Jan 21, 2014
  1. Detecting DISPLAY_DEGREE in list not done correctly, so when

    DISPLAY_DEGREE was passed it was ignored and reset to 0 (fixed).
  2. Using crs.messages=FALSE to set opts$DISPLAY_DEGREE=0 unless

    DISPLAY_DEGREE is an argument. This will disable console I/O as
    desired (also have updated the FAQ)
Commits on Jan 20, 2014
  1. Corrected glitch with print.output not being set when

    crs.messages=FALSE (ought to have been set to FALSE)
Commits on Jan 19, 2014
  1. Reinserted eval.lscv/eval.aicc in glpcvNOMAD to avoid warning about

    ... Removed kernel as a return argument in npglpreg as it was not
    returned (left over from crs)
  2. Cleaned up lingering warning about <anonymous>:... caused by not

    feeding optional args ... all the way through the codebase
Commits on Jan 18, 2014
  1. In npglpreg cv.maxPenalty is now a passable parameter (was

    sqrt(.Machine$double.xmax) but could have potentially been used
  2. Moved functions eval.lscv and eval.aicc outside of glpcvNOMAD to

    streamline code. Modified default to issue warning when degree.max is
    beyond the maximum well-conditioned basis possible (was forcing me to
    stare at code wondering why starting values did not lie uniformly in 0
    degree.max when I set degree.max to 100)
  3. Added detail on optimization and cautioning users to investigate the

    impact of tuning parameters for snomadr()
Something went wrong with that request. Please try again.