2023-07-12 Scott Bartell
* DESCRIPTION (Version): New version is 1.3
* R/trimdata.R and R/predgrid.R: revised functions to rely on functions from sf,
rather than PBSmapping and maptools, to prepare for upcoming deprecation of PolySet
class in maptools. Newly revised functions can now handle a broader set of map classes,
including: map, sf, Raster (with some limitations), and Spatial.
* updated various help file examples to remove optional use of PBSmapping, because the
necessary functions (now only used in sampcont.R) are automatically imported from that
package. References and URLs updated in several help files.
2022-05-25 Scott Bartell
* DESCRIPTION (Version): New version is 1.2-6
* R/colormap.R, R/trimdata.R, and R/predgrid.R: replaced class checks for different
map types with inherits function instead of using ==, as per CRAN recommendation.
* R/modgam.R and R/summary.modgam.R: Now only produces a global p value if the number
of rows in the model fits are identical for the smoothed model and the null model (i.e.,
there are no missing data in the variables being smoothed). See modgam help file for
more information.
* R/sampcont.R: Fixed errors that had caused nrow and ncol values to be swapped,
reversing the intended dimensions for the stratified sampling grid. Also now added the
sampling grid as output from the function, in order to facilitate plotting. Thanks to
Ian W. Tang, University of California, Irvine, for discovering the problem and providing
code for these changes.
2019-07-27 Scott Bartell
* DESCRIPTION (Version): New version is 1.2-5
* R/mypredict.gam.R: Fixed error in mypredict.gam function associated with change in
default argument in vcov function since R 3.5.0.
* R/modgam.R and R/plot.modgam.R: Added "pointwise" argument with default value FALSE,
to skip pointwise permutation tests unless requested by the user. Identification of
high risk regions is better accomplished using se.fit=TRUE, available since version
1.0. Plot function warning message updated to reflect new argument.
2019-01-21 Scott Bartell
* DESCRIPTION (Version): New version is 1.2-4
* R/plot.modgam.R: Fixed error in plot.modgam function associated with new "anchor"
argument, now implemented in both plot.modgam and colormap.
* R/gamcox.fit.R: Truncated step sizes (deltaeta) at -3 and 3 in order to avoid
potential problems. Also added warning message when final model deviance is higher
than deviance for a model using the same linear predictors and no smooth, and error
check for infinite values in eta.
* R/optspan and R/gamcox: Fixed same error in both functions--had sometimes
failed to identify the variables in the two-dimesional smooth, occasionally
resulting in fatal error messages.
* R/gamcox: replaced identical with all.equal to avoid spurious warning/error
when comparing two real values (span and spanf).
2018-07-24 Scott Bartell
* DESCRIPTION (Version): New version is 1.2
* R/colormap.R: fixed error with double printing of y axis label when axes=TRUE,
and added a new argument "anchor", with a default of anchor=F (the original method).
Anchoring the color scale bar to the referent value is now only performed when
legend.add.line is specified and anchor=T.
2018-04-09 Lu Bai and Scott Bartell
* DESCRIPTION (Version): New version is 1.1
* R/summary.modgam.R: changed class check from "gam" to "Gam" to accomodate class
name change from the gam library (starting with gam version 1.15).
* R/mypredict.gam.R: corrected error in dispersion parameter estimation for
quasibinomial and quasipoisson families. Special thanks to Fabian Camacho,
University of Virginia, for noticing the problem.
* R/colormap.R, R/plot.modgam.R: several modifications: new arugment allows
users to specify any color palette, default palette for plot.modgam is changed to
a divergent palette, new colormap help file example compares palettes, the color
palette is now anchored on legend.add.line (if legend.add.line is specified),
the color gradient now changes on an exponential (instead of linear) scale when
exp=T, help file now suggests readOGR instead of readShapePoly for importing
shapefiles, and formatting changed in the legend.
* added more S3 methods (coef(), formula(), summary(), AIC(), residuals())
* copyright license information (GPL-3) added to all R code files
* other minor changes and bug fixes
2016-06-19 Lu Bai
* DESCRIPTION (Version): New version is 1.0
* R/gamcox.R, R/print.gamcox.R, R/predict.gamcox.R: MAPGAM now includes functions
to fit a Cox additive model for survival data by calling gamcox or specifying
family = "survival" in modgam. The model returned a object of class "gamcox",
which can be expressed by print and predict.
* R/modgam.R: modgam now can fit both GAM and GAMCOX models. Added an option
"survival" for argument "family". Added "formula" argument(optional), which
allow for specifying the model using formula expression. Added "se.fit" argument
to specify whether to compute pointwise standard errors and confidence intervals.
Set a class for the returned object, "modgam". The returned object has some
differences from previous versions of MapGAM, such as "exp" instead of "OR" for
exponentiated fit values.
* R/mypredict.gam.R: Can be used to compute pointwise standard errors for new
predictions based on a gam object which the predict.gam function in gam package
can not do.
* R/optspan.R: Added "spans" argument so that the candidate spans can be specified
flexibly.
* R/colormap.R: Added an option "interval" for argument "contours" to plot countours
to specify areas where confidence intervals exclude zero. Added "contours.drawlables",
"contours.lwd", "contours.lty", "contours.levels", "contours.labcex" to make plotting
contour lines more flexibly. Added "legend.name", "legend.cex" and "legend.add.line"
to improve appearance.
* R/plot.modgam.R, R/print.modgam.R: print and plot functions for object returned
by modgam. For plot, optionally, if the object containes confidence intervals,
colormap of confidence intervals can be generated.
* data/CAdata.rda, data/CAgrid.rda, data/CAmap.rda: Added three data set for examples
of analysing spatial effects on survival rates.
2015-05-02 Scott Bartell
* DESCRIPTION (Version): New version is 0.7-5
* R/colormap.R: added "axes" argument, which improves appearance and readability
of chemical mixture interaction plots when set to TRUE.
* R/trimdata.R: changed CRS function call to conform to new error checks in the
sp package.
2014-12-10 Scott Bartell
* DESCRIPTION (Version): New version is 0.7-4
* R/colormap.R: changed to improve scale size and appearance, and position
of legend. Added "contours" argument for contours based on either permutation
test p-values or predicted response. Also fixed error in example code.
* NAMESPACE: added import and importFrom commands to conform to new CRAN policy.
* R/sampcont.R, R/trimdata.R: added PBSmapping:: to functions called from
that package, and updated documentation to clarify which options require
PBSmapping.
2013-12-12 Scott Bartell
* DESCRIPTION (Version): New version is 0.7-0
* R/modgam.R, R/optspan.R, R/colormap.R: MapGAM now includes an
unconditional permutation test, and allows the user to select
any family and link function. The package was previously limited
to the (now default) binomial family with a logit link, but potential
applications are broader such as investigation of joint effects of chemical
mixtures (suggested by Tom Webster, Boston University). The relevant
help files are updated to reflect these changes, and new examples have
been added to the modgam help file.
* data/MAdata.rda: Added two new columns to the simulated data set,
for use in a new chemical mixtures example.
2013-03-11 Scott Bartell
* DESCRIPTION (Version): New version is 0.6-2
* DESCRIPTION (Depends/Suggests): PBSmapping changed from Depends
to Suggests, for compatibility with new PBSmapping license.
* man/trimdata.Rd, man/predgrid.Rd, man/MapGAM-package.Rd,
man/colormap.Rd, man/sampcont.Rd: only test certain examples if
PBSmapping is installed, and note when non-tested examples require
PBSmapping.
2012-12-23 Scott Bartell
* DESCRIPTION (Version): New version is 0.6-1
* R/modgam.R (modgam): corrected error in p-value calculation for
global permutation test for significance of location. Thanks to
Lucheng Shao, UC Irvine, for noting strange simulation results.