2013-07-12 pauljohn
* ../R/plotPlane.R (myPredict): re-work so user can supply
arguments x1lim, x2lim, and the subsidiary work narrows attention
to those observation ranges. Careful to re-work dot argument
processing and block users from supplying xlim, ylim, directly.
* ../R/perspEmpty.R (perspEmpty): correct spacing around equal signs
2013-07-01 pauljohn
* ../R/residualCenter.R (predict.rcreg): change predict.rcreg
because of problem with newdata going into dots when it is not
wanted there.
2013-06-24 pauljohn
* ../R/utils.R (cutBySD): add n = 2 definition for cutBySD and
cutByQuantile methods.
* ../R/meanCenter.R: runtime error from predict.mcreg because the
argument newdata was double-entering through ...
2013-06-06 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): col needs to be in arguments
to plotFancy. Same for plotCurves
(plotFancy): color name problem when user supplies col argument. Fixed.
* ../R/genCorrelatedData.R (genCorrelatedData2): recycle sds
argument to length P
* examples/plotSlopes-ex.R, plotCurves-ex.R. Incorporate
genCorrelatedData2.
2013-06-02 pauljohn
* ../DESCRIPTION (Date): New data set: Nyblade-Read
cheating/looting data for Japanese politics.
2013-05-31 pauljohn
* ../R/predictOMatic.R (predictCI): futzing about to deal with
problem in predictCI. If interval="none", we should just take
the model's predict output and format it, assuming it is a vector
or matrix. The other messing about with intervals is only needed
if the user asks for that, and so more regression types will work
with predictOMatic this way.
2013-05-24 pauljohn
* ../R/testSlopes.R (plot.testSlopes): Continued long battle with
too confusing system for shading and labeling intervals. This is a
big example of how expedient solution lead to bad code, hard
to bug-fix, frustrating as hell. Now see variable markAt, which
tries to simplify one part. Still, rest not great.
2013-05-22 pauljohn
* ../R/plotPlane.R (myPredict): embarrassing: noticed that I had
written an argument "linesFrom" in plotPlane and had forgotten
that and re-wrote same functionality in addLines(). So now came
back to re-work linesFrom so it causes same output as addLines.
2013-05-21 pauljohn
* ../R/plotSlopes.R (plotFancy): llwd colors were not being
passed to plotFancy and they were not being used by lines().
2013-05-19 pauljohn
* ../R/plotCurves.R (plotCurves): found bug in dotargs sorting for
predict, Was causing crash in logit models
* ../R/testSlopes.R: format spaces
* ../R/residualCenter.R (predict.rcreg): format spaces
* ../R/mcGraph.R: format spaces and T -> TRUE
* ../R/centralValues.R (centralValues): Style cleanup, T -> TRUE
* examples/predictOMatic-ex.R: range(x) -> range(x, na.rm = TRUE)
to allow missings
* ../R/predictOMatic.R (predictOMatic): se.fit: try not to
return it unless predict argument se.fit = TRUE is passed.
2013-05-16 Paul Johnson
* ../R/genCorrelatedData.R (mvrnorm): insert repaired code for
mvrnorm and use that in genCorrelatedData functions.
(genCorrelatedData2): Using lower triangular metaphor for beta2.
(genCorrelatedData2): inserted some verbose printing for the equation
2013-05-15 pauljohn
* ../R/outreg.R (outreg2HTML): fixing error with multicolumn
titles when tight = F
* examples/testSlopes-ex.R: spacing and indentation
2013-05-15 Paul Johnson
* ../R/plotSlopes.R (plotFancy): plotFancy is the new backend
drawer for plotSlopes and plotCurves
* examples/plotCurves-ex.R: clean up formatting, spaces around
equals, after commas
2013-05-14 pauljohn
* ../R/predictOMatic.R (lenVar): Change from length to NROW
because that gives true answer for vectors or matrices, and some
models are going to have matrix DV data (grouped binomial)
* examples/model.data-ex.R: put no-data.frame examples at top.
Fix typos in examples.
* ../R/predictOMatic.R (lenVar): Threw away model.data approach.
Realized I can manufacture formula from variable names and then
model.frame can generate the data frame I want from that.
2013-05-13 pauljohn
* ../R/utils.R (cutByQuantile): change cutByQuantile to 0, 25, 50,
75, 100.
* ../R/plotSlopes.R (plotSlopes.lm): depVar: trying to
eliminate assumption this is only numeric.
2013-05-12 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): try to grab the
predict-intended arguments out of the dotargs.
* Use predictCI to calculate predictions.
* ../R/predictOMatic.R (predictOMatic): predictCI new function
tries to give c(fit, lwr, upr) for lm and glm. Maybe others.
* tie that into predictOMatic, so interval can be returned.
* tie that into plotSlopes, so glm interval can be plotted.
2013-05-09 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): predict and dotargs.
2013-05-07 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): opacity new argument.
2013-05-06 pauljohn
* ../R/meanCenter.R (std): noticed that meanCenter did not work
with glm, and don't see why it shouldn't, as long as centerDV is
handled properly.
* ../R/predictOMatic.R (predictOMatic): had duplicate usage of
predict that caused it to fail.
2013-05-05 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): yet still more futzing about
with color matching of factor levels, points, lines, and regions.
2013-05-04 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): color fix for user=specified
color matrices
2013-05-03 pauljohn
* examples/plotSlopes-ex.R: add prestige interaction at end.
Used dev.new(). hope that's not a violation.
* ../R/plotSlopes.R (plotSlopes.lm): Yet more battle to match
factor and numeric modx colors and line types.
TODO: llwd still problematic
fix plotCurves.R similarly
2013-05-02 pauljohn
* ../R/plotSlopes.R (plotSlopes.lm): frustrating problem with
colors assigned to points, lines, intervals, and legend items
when user selects just a few values for plotting.
* ../R/meanCenter.R (predict.mcreg): The drop gotcha again.
tmeans <- sapply(newdata[ , nc], mean, na.rm=T) must be:
tmeans <- sapply(newdata[ , nc, drop = FALSE], mean, na.rm=T)
2013-04-30 pauljohn
* ../R/outreg.R (sigtest): nobs -> stats::nobs because gdata had
its own nobs
2013-04-29 pauljohn
* ../R/outreg.R (outreg2HTML) had to be exported.
* (outreg) new outreg, that creates a character vector
and cats it to screen, and returns it as well.
* (outreg2HTML): new function that can write an html
file.
2013-04-27 pauljohn
* ../R/outreg.R (checkReg): check if model has methods summary,
coef, vcov, and nobs
* outreg: be more careful receiving modelList, checking if it
is a regression model.
2013-04-24 pauljohn
* ../R/testSlopes.R (plot.testSlopes): fix typo in which numeral 2
was replated by Tcrit in bsse calculation
* ../R/predictOMatic.R (newdata): fixing predVals declaration
* outreg: multiple alpha values
2013-04-23 pauljohn
* ../R/outreg.R: found bug where df.residual was not
correct for each separate model. fixed!
2013-04-22 Paul Johnson
* ../R/utils.R (getFocal: create separate documentation for
the methods, try to clarify usage
(cutBySD): allow m+/-3sd
2013-04-21 pauljohn
* ../R/predictOMatic.R (focalVals): allow divider to be a
function, such as R's built-in pretty() function.
- create new argument "predVals" to replace fl
* Version 1.7.4.1
* ../R/genCorrelatedData.R: Added whole family of new
funcions
- genCorrelatedData2
- vech2Corr
- vech2Mat
- lazyCor
- lazyCov
- makeVec
- makeSymmetric
- checkPosDef
* ../R/mcDiagnose.R (getDeltaRsquare): drop = FALSE should return
a matrix with one column
- more revision on getPartialCor
2013-04-19 pauljohn
* ../R/mcDiagnose.R (mcDiagnose): getPartialCor function added.
2013-04-18 pauljohn
* ../R/predictOMatic.R (predictOMatic): fiddling interface for fl
so it can now be offered as a vector or a list. Trying to get cute
with way users can specify divider algorithms.
* ../R/utils.R: must now export the cutBy functions
2013-04-17 pauljohn
* ../R/mcDiagnose.R (getDeltaRsquare): clean up output. Because
I was using na.omit to get rid of the missing first element (some
noise from matrix subtraction), the output object was including a
lot of extraneous attributes that made printing the result more
tedious. Now I don' use na.omit any more.
2013-04-16 pauljohn
* ../R/outreg.R (harvest): Again reworking arguments, rename
extraSums to "request" and extraFuns to "runFuns". More examples.
- new arguments, alpha.
* ../R/predictOMatic.R (focalVals): new function, a wrapper
around divider functions.
* outreg: renaming arguments
2013-04-15 pauljohn
* ../R/predictOMatic.R (predictOMatic): introduce "AUTO"
keyword for focus list variables.
* ../R/outreg.R (gofPrint): rewrite to add extraFuns and
extraSums. Mostly working.
(outreg): fixing logical flaw when a model has NO summaries we
want. Fighting with nonNull attr
2013-04-14 pauljohn
* ../R/outreg.R (outreg): Reformat consistent with Rstyle
- change lyx = TRUE argument to float = FALSE.
(outreg): reworked to grab B and SE matrices, create PT
at top.
2013-04-10 pauljohn
* ../R/testSlopes.R (plot.testSlopes): because of mis-match
between roots and edges of MMlwr and MMupr, re-designed this
to calculate interval types and ranges exactly, iterate
over intervals to shade appropriately. Still Ugly code,
only slightly less embarrassing.
2013-04-09 pauljohn
* ../R/testSlopes.R (plot.testSlopes): Fixing MMupr and MMlwr
selection and thinning.
2013-04-08 pauljohn
* ../R/testSlopes.R (plot.testSlopes): add shade argument to
stop the pink shading. TODO. allow user color choice.
* Found Bug on MMlwr and MMupr: don't exclude middle
in two-ended significant interval. Must think harder.
* ../R/utils.R (cutByQuantile): now allow n=2 imply quantiles 25
and 75.
* examples/testSlopes-ex.R: new example for testSlopes, cut
example code from testSlopes.R
* ../R/testSlopes.R (testSlopes): requires plotSlopes object
* ../R/plotSlopes.R (plotSlopes.lm): class now "plotSlopes" and
"rockChalk". plotPlane works with all 2d rockchalk objects, but
testSlopes only works with plotSlopes so far
2013-04-02 pauljohn
* examples/meanCenter-ex.R: use new "terms" argument to specify
variable to be centered.
* ../R/meanCenter.R (print.summary.stdreg): meanCenter=add
terms argument to allow user to circumvent the automatic
variable selection and force the centering to be done.
2013-03-31 pauljohn
* Examples: plotCurves-ex, plotSlopes-ex: show new features
* ../R/testSlopes.R (testSlopes): pass inroot into jn, so the plot
function can check it.
* ../R/plotCurves.R (plotCurves): now accepts modx as option,
will plot without it.
* legend and color re-fitted. Nearly going insane maintaining
almost identical plotSlopes and plotCurves. Must regain sanity.
* ../R/plotSlopes.R (plotSlopes.lm): clean up color
access, fix legend
2013-03-29 pauljohn
* ../R/plotPlane.R (addLines): saved the col value in the
return of plotSlopes. Use that in addLines unless user said
otherwise in col argument.
* ../R/plotSlopes.R: working through assinine decisions I
carelessly made on changing handling of interval in plotSlopes.
Now behavior slightly more predictable. But frustating.
2013-03-28 pauljohn
* ../R/plotSlopes.R (plotSlopes): revise to allow missing or null
modx.
* note bug in color assignment for points with factors. Fixed.
2013-03-27 pauljohn
* ../R/testSlopes.R (testSlopes): separate calculation of
testSlopes data from plotting and printing methods.
(plot.testSlopes): plot.testSlopes is sorta-working now,
not cleaned up.
* massively clean up logic, eliminate nested loops by
returning when phase of calculation is finished.
* fix VERY STUPID variable names that have the dash
in them. "hypo-test" was just dumb.
2013-03-26 pauljohn
* ../R/utils.R (cutBySD): Changed "cutNumeric" and "cutFactor" to
a generic function "getFocal" with methods getFocal.default and
getFocal.factor.
(getFocal.factor): convert to if/return stanzas.
* predictOMatic.R (model.data): silencing warning
about re-used row names. Warning was triggered by
this "data <- data.frame(ldata[varNames])" "Warning in
data.row.names(row.names, rowsi, i) : some row.names duplicated:
3,5, --> row.names NOT used" This was a harmless warning, but
bothersome to users.
* Rename rnOrigData -> origDataRN
2013-03-24 pauljohn
* ../R/utils.R (cutNumeric): New function to integrate work that
was duplicated in many separate functions, plotSlopes, plotCurves
* ../R/plotSlopes.R (plotSlopes): interval argument
* Use new cutNumeric function to find interesting values in a
numeric plotx
* plot sub-routine for shaded pred or conf intervals
* incorporate newdata() function, replace ad-hoc code
* n argument to feed to newdata()
2013-02-19 pauljohn
* ../R/testSlopes.R (testSlopes): Sunthud Pornprasertmanit
suggests 2013-02-19 there is a bug, diag is applied to a singleton.
2013-02-17 pauljohn
* ../R/plotPlane.R (myPredict): class set to rockchalk3d
* ../R/plotSlopes.R (plotSlopes): class set to rockchalk2d.
* ../R/plotCurves.R (## carrier.name): class set to rockchal2d.
* ../R/utils.R (cutByTable): make sure return value is a factor
for factor x, and numeric for numeric x
2013-02-16 pauljohn
* ../R/predictOMatic.R (predictOMatic): return NOT invisible()
anymore. Why did I do that at before?
* ../R/plotSlopes.R (plotSlopes): removed moderator variable name
from legnd variable, after adding variable name to legend title
* ../R/utils.R (cutByTable): Fixed the longstanding bug that
cutTable did not return the desired values (had names and
values reversed in role). Now output same meaning as
cutByQuantile
2012-06-03 pauljohn
* ../R/plotPlane.R (## carrier.name): using model.data now, and
varNamesRHS, no need for carrier nonsense anymore.
2012-06-02 pauljohn
* ../R/plotCurves.R (carrier.name): revise to use model.data
function, including the attribute varNamesRHS. Cleans up use of
carrier data.
2012-05-29 pauljohn
* ../R/meanCenter.R (formulaReplace): rework code so centered
variables are re-named with "c" on end. Otherwise too confusing.
* standardize: re-work to change variable names by appending s.
* centerNumerics: new function, will add centered variables to a
data frame with new names with "c" or "cs" appended.
2012-05-28 pauljohn
* ../R/meanCenter.R (predict.mcreg): predict must use same mean
and scale as were used in original regression. Use meanCenter
centeredValues return more fully.
* meanCenter(std): eliminate use of R scale, now calculate
mean and scale factors, return them in list with centered
variables.
2012-05-27 pauljohn
* ../R/plotPlane.R (carrier.name): use model.data, fix centerValues
* ../R/residualCenter.R (createRCinteraction): Noticed that this
function did not work with categorical variables in
interactions. MASSIVELY reworked calculation approach. Because R
generates new variables when it finds factors, a new, completely
literal approach is taken. From fitted model, get the design
matrix, notice the colnames are the variables in a linear
formula. Scan through there and replace variables with colons with
the residual centered versions.
2012-05-19 pauljohn
* ../R/recodeFactors.R (adjacent): fix flaw in ordered factor
re-calculation (had *again* made the "levels" mistake on the RHS)
2012-05-13 pauljohn
* ../R/plotPlane.R: x1floor and x2floor. Bad names? Number of
divider lines to draw in floor of plot. Default 5, which was the
value I had hard coded before.
* ../R/perspEmpty.R (perspEmpty): clean up interface! Use x1 and
x2 arguments to position the bottom grids. Will require
interface cleanups in other functions.
* ../R/mcGraph (mcGraph1-3): Changes to make usage match
perspEmpty intercface change. Instead of assuming perspEmpty will
manufacture the plot sequences, now the caller must do so.
2012-05-07 pauljohn
* ../R/plotPlane.R (myPredict): move draw arrows before points, so
arrows sit under, not over points
* ../R/mcGraph.R (mcGraph3): name elements in returned list
2012-05-06 pauljohn
* ../R/plotSlopes.R (plotSlopes): llwd: lwd to pass through to
lines. lwd controls darkness of points only.
* ../R/outreg.R (outreg): fix conditional at end to insert
\end{table} to match previous changes to top of function
2012-05-05 pauljohn
* ../R/magRange.R (magRange): work to allow mult smaller than
1. Range can shrink to (center, center) point when mult reaches 0.
2012-04-29 pauljohn
* ../R/summarize.R (summarizeNumerics): insert N output for summarize
2012-04-28 pauljohn
* ../R/utils.R (cutByQuantile): work out n to divide interval
according to a special scheme
(cutBySD): choose between n=3 and 4
* ../R/summarize.R (summarizeFactors): if no factors found, return NULL
(summarizeNumerics): if no numerics, return NULL
2012-04-20 pauljohn
* ../R/summarize.R: sumstat: new paramter to restrict diagnostic output
(summarizeNumerics): fix crash when data frame does not includes
only factors or only numerics. Returns NULL in summarizeNumerics
or summarizeFactors.
2012-04-15 pauljohn
* ../R/meanCenter.R (standardize.lm): drop=FALSE for dm, keeps it
a matrix
(summary.mcreg): drop=FALSE for dm
2012-04-04 pauljohn
* ../R/testSlopes.R (testSlopes): fixing plots. red intervals on "either
side" result. Labels too.
2012-04-02 pauljohn
* ../R/plotCurves.R (cutBySD): insert "col" option, similar to plotSlopes
2012-03-31 pauljohn
* ../R/plotSlopes.R: allow user to specify colors in plot. add
option plotLegend.
* ../R/outreg.R (outreg): fixing problem that title did not print
unless LyX=FALSE
2012-03-29 pauljohn
* ../R/summarize.R (ncw): new method print.factorSummaries
2012-03-09 pauljohn
* examples/meanCenter-ex.R: remove references to centerContrasts
* ../R/meanCenter.R (std): predict.mcreg centers all variable in the
"centeredVars" list.
* ../R/meanCenter.R: Cut "centerContrasts" option in meanCenter and tons of
beautiful code because it is impossible to manage re-centering in
predict.mcreg if there are centered congrasts.
2012-03-03 pauljohn
* examples/predict.rcreg-ex.R: new file to demo predict
* ../R/residualCenter.R (predict.rcreg): new method handles
by re-calcuating predictions for each auxiliary regression.
2012-03-01 pauljohn
* ../R/meanCenter.R (std): Found really bad bug where
plotPlane(meanCenter) returned error about predict being
used for wrong type of data. Problem traced back to fact
that "scale" returns a matrix with 1 column, rather than
a numeric vector, so dataClasses was mismatched.
attr(,"dataClasses")
y xxx x22
"numeric" "nmatrix.1" "nmatrix.1". The "newdata" I'd
generate did not have "nmatrix.1" as the type, and so
predict's use of .MFCheckClass resulted in failure.
Like others, I say "How aggravating."
2012-02-01 pauljohn
* ../R/plotSlopes.R (plotSlopes): "std.dev." and "quantile" now
legal strings for modxVals, along with particular values.
2012-01-30 pauljohn
* ../R/plotPlane.R (plotPlane.default): correct typos referring to plotSlopes
* ../R/plotSlopes.R (plotSlopes): add na.rm=TRUE for range commands
2012-01-25 pauljohn
* ../R/plotSlopes.R (plotSlopes): change return value to include newdf
2012-01-09 pauljohn
* ../R/lmrc.R (makeRCformula): weird. data from residuals turns up
with class "AsIs" and causes all kinds of confusion later
* residualCenter now a generic method
* residualCenter.default should work for any glm or lm.
2012-01-03 pauljohn
* ../R/meanCenter.R (std): create class name for mean centered
regression, add attributes.
print.mcreg: new method
2011-12-30 pauljohn
* ../R/plotPlane.R (carrier.name): use get_all_vars to build emf
2011-12-28 pauljohn
* ../R/plotPlane.R: insert plotPoints option to suppress plotting
of points. Question. Should plotPoints=FALSE force drawArrows=FALSE?
2011-12-12 pauljohn
* ../R/plotSeq.R: allow x to be a factor, return levels if so
* /pauljohn.plotSeq.R (plotSeq): range
allow missings
* ../R/plotPlane.R: new introduced function
2011-12-11 pauljohn
* ../DESCRIPTION (Date): Version: bump
* ../R/perspEmpty.R (perspEmpty): range(na.rm=TRUE) allow missings
* ../R/magRange.R (magRange): range(na.rm=TRUE) to allow missings
2011-12-06 pauljohn
* ../R/lmrc.R (makeRCformula): This is necessary so that all
interactive components are included
iv <- paste("(", gsub(":"," + ",x),")^2")
* Replace whole stucture to make sure x1.X.x2 coding gets used.
* ../R/plotSlopes.R (plotSlopes): trying to fix that pesky "one
extra variable" error in regression models
2011-12-02 pauljohn
* ../R/plotSlopes.R (plotSlopes): fix problem when only one
variable was left in calculating centralized predictor values. The
bare vector has to be told "as.data.frame"
2011-12-01 pauljohn
* ../R/lmrc.R (makeRCresiduals): revised bug in calculation lmrc
was not using residual centered variables when "x1:x2" was
included, R was constructing own product term. So necessary to
re-name the residual centered variable connector .X.
2011-11-26 pauljohn
* ../R/mcGraph.R (mcGraph3): added option drawArrows
2011-11-05 pauljohn
* ../R/mcGraph.R (mcGraph3): insert invisible() on returns
* ../R/rockchalk-package.R: insert dataset religioncrime and
documentation.
* data/religioncrime.RData: new dataset with beautiful example
of suppression
2011-11-02 pauljohn
* rockchalk/rockchalk/R/mcGraph.R (mcGraph3): more fiddling
with interaction
* mcGraph.R (mcGraph3): fix interaction code in lm