--------------------
version 2.8
-----
* In functions ``ffRanefLMER.fnc'', disabled the blocking
of random effects that are not in the fixed effect structure.
This was done so as to enable the addition of random effects
such as ``(1+VAR1*VAR2|VAR3)''.
--------------------
version 2.5
-----
* Fixed a bug win ``bfFixefLMER_F.fnc'' and
``bfFixefLMER_t.fnc'' when backfitting on AIC,
BIC, llrt, relLik.AIC, and relLik.BIC.
* Added option "z" to argument ``method''.
* Added examples to test backfitting on all methods for
lmer and glmer models.
--------------------
version 2.4
-----
* Updated package to work with post-version 1.0 of
package lme4.
* Revived, so to speak, function ``plotLMER.fnc'' from
archived package ``languageR'' and revised everything to
be compatible with lme4 version > 1.0.
* Removed any mention of ``pvals.fnc'' and associated
functions from archived package ``languageR'' from any
function that used it.
* In functions ``ffRanefLMER.fnc'' and ``fitLMER.fnc'',
argument ``data'' is depricated.
--------------------
version 2.1
-----
* Updated the percentage of deviance explained in function
``pamer.fnc''.
* Revised the random effects backfitting algorithm:
If you have the model
lmm.1<-lmer(abundance~year+(year|site))
where year is not significant, then
-- old backfit:
- year is removed from fixed effects
- random effect is completely removed (which can actually
be problematic (statistically), if it is part of your
experimental set up, leading to pseudoreplicates).
- problems with lmer because no random effect in formula
-- new backkfit:
- year is removed from fixed effects
- random effect is adjusted. new formula: abundance~(1|year)
- no problems with lmer because no random effect in formula
--------------------
version 2.0
-----
* In functions ``bfFixefLMER_t.fnc'', backfitting when the
model has factor variables with levels greater than 2 will
issue a warning (instead of stopping the process).
* Added backfitting on AIC, BIC, relLik.AIC, or relLik.BIC
capabilities to functions ``bfFixefLMER_F.fnc'',
``bfFixefLMER_t.fnc'', and ``fitLMER.fnc''.
* Also added functions ``bfFixefLMER_F.fnc'',``bfFixefLMER_t.fnc'',
and ``fitLMER.fnc'' an option to not test (and thus not remove)
main effects.
* Updated the arguments in function ``fitLMER.fnc'' accordingly.
* In function ``ffRanefLMER.fnc'', changed the default of argument
``log.file'' to ``NULL''.
* Fixed a bug in fucntion ``bfFixefLMER_t.fnc'' when
backfitting glmer models.
* Function ``ffRanefLMER.fnc'' can now be used with glmer models.
* Fixed a bug in function ``fitLMER.fnc'' regarding the forward
fitting of random effects.
* WISH LIST: AIC, BIC, relLik.AIC, and relLik.BIC for
random effect forward fitting (for function
``ffRanefLMER.fnc'').
--------------------
version 1.7.0
-----
* Fixed little bug in functions ``bfFixefLMER_t.fnc'' and
``bfFixefLMER_F.fnc'' regarding the pruning of random effects.
* Put back a piece of code that was mistakingly removed in
previous version in ``bfFixefLMER_t.fnc''.
* When the model is a GLMER, ``z-value'' instead of
``t-value'' is printed.
* Added function ``relLik''.
* Added argument ``verbosity'' to function ``mcposthoc.fnc''.
* Added function ``summary.mcposthoc''.
* In ``plotLMER3d.fnc'', ``plotDensity3d.fnc'', and ``plotRaw3d.fnc'',
for 3d perspective plots (``persp3d'') changed the edges of the color
matrix from 0 to 1.
* Re-vamped examples to make them run faster.
* Added comments to code.
* Function ``posthoc.fnc'' now depends on package ``parallel''.
* In ``plotDensity3d.fnc'', ``plotLMER3d.fnc'', and ``plotRaw3d.fnc'',
removed argument ``ret''. now invisibly returns potting information.
* Added plot.type = ``image.plot'' to functions ``plotLMER3d.fnc'' and
``plotRaw3d.fnc''.
--------------------
version 1.6.8.3
-----
* Fixed a bug in ``bfFixefLMER_F.fnc'': The script did not
update the model after removing the last parameter within
a specific order. Now it does.
* In functions ``bfFixefLMER_t.fnc'' and ``bfFixefLMER_F.fnc'',
prior to this change, it was absolutely necessary that names of
factors should not occur as parts of other factor namess, e.g. "c"
and "cv". The script would recognise presence of "c" whereever "cv"
is present. Now it looks for an exact string match, so factor "c"
will not be found in factor "cv".
* Added glmer capabilites for function ``mcposthoc.fnc''.
--------------------
version 1.6.8.2
-----
* Fixed an issues in ``bfFixefLMER_t.fnc'' with keeping
non-significant factors when they are present in higher
order interaction terms. For example: A non-significant
interaction term "c:d" will be excluded even though "c:t:d"
is significant. Thanks to Johannes Ransijn
for this one.
* Function ``posthoc.fnc'' is now depricated and replaced
by ``mcposthoc.fnc'';
* Added MCMC p-values functionality to function ``mcposthoc.fnc'';
* Shortened example run time;
* Removed argument ``data'' from functions ``bfFixefLMER_F.fnc'',
``bfFixefLMER_t.fnc'', and ``fitLMER.fnc''.
--------------------
version 1.6.7
-----
* Added arguments ``add.raw'', ``color.raw'', and ``alpha.raw'' to
function ``plotLMER3d.fnc'', and fixed a little bug in that function.
* Removed argument ``ret'' from function ``plotLMER3d.fnc''. The plotting
data is now invisibly returned.
* Removed dependency on package ``languageR''.
* Fixed a little bug in function ``ffRanefLMER.fnc''.
* Removed all data sets and put them in a new package called ``LCFdata'',
which is now suggested.
* Added argument ``scale'' to function ``plotRaw3d.fnc'' and
``plotLMER3d.fnc''.
* Added argument ``shift'' to function ``plotLMER3d.fnc''.
* Changed default value of argument ``shift'' in function ``plotRaw3d.fnc''
from NULL to 0.
--------------------
version 1.6.6
-----
* Removed suggestion to package ``Design''
and changed examples accordingly.
--------------------
version 1.6.4
-----
* Fixed little bug in function ``ffRanefLMER.fnc''.
* Added argument ``if.warn.not.add'' to functions ``ffRanefLMER.fnc'' and
``fitLMER.fnc''.
* Added arguments ``plot.dat'' and ``path'' to function
``plotLMER3d.fnc''.
* Added arguments ``add'', ``alpha'', and ``xy'' to function
``plotRaw3d.fnc''.
* In function ``plotLMER3d.fnc'', can now plot even if error message
``Error in persp3d.default(x = as.numeric(rownames(z)),
y = as.numeric(colnames(z)), : increasing 'x' and 'y' values expected''
occurs. Issues the following warning: ``plotting anyways, but will not
use supplied x- and y-values ...''.
* Added model ``z'' to run a new example in function
``plotRaw3d.fnc''.
* Added function ``mcposthoc.fnc''.
--------------------
version 1.6.3
-----
* In function ``plotLMER3d.fnc'', added the possibility to
add a tittle to "persp" plots, in addition to
* the option of having a flat miror image underneath the main 3d
surface of "persp3d" plots,
* rug plots for the three types of plots and for the flat
miror image underneath the main 3d surface of "persp3d" plots,
* transparency control for the main 3d surface, reference level,
and flat miror image for "persp3d" plots, and
* the option to return the z and color matrices used for
plotting.
* Added function ``plotDensity3d.fnc''.
* Added function ``plotRaw3d.fnc''.
* Added data ``erpFz''.
--------------------
version 1.6.2
-----
* Added argument ``data'' to function ``posthoc.fnc''.
* Added argument ``prior.releveling'' to function ``posthoc.fnc''.
--------------------
version 1.6
-----
* Added function ``plotLMER3d.fnc'' to create (dyanmic) 3d
plots of models fitted with LMER.
* Added sub-argument ``corr'' to argument ``ran.effects'' in
functions ``ffRanefLMER.fnc'' and ``fitLMER.fnc''.
* In function ``ffRanefLMER.fnc'', when ``ran.effects'' is a vector
(e.g., c("(Condition | Subject)", "(0 + TrialNum | Subject)")), the
function now checks whether variables specified in ``ran.effects''
are present in the fixed-effects structure. If the variable appearing
in a ranef is not present in model coefficients, it is skipped.
* Fixed a little bug in function ``ffRanefLMER.fnc'' regarding
the testing of random effects of the form ``(Condition|Subject)''.
* Fixed a little bug in functions ``bfFixefLMER_F.fnc'' and
``bfFixefLMER_t.fnc'': When all model terms are removed, the
functions now return a model with only an intercept and whatever
random effects were in the model.
* Added argument ``prune.ranefs'' to functions ``bfFixefLMER_F.fnc'',
``bfFixefLMER_t.fnc'', and ``fitLMER.fnc''.
--------------------
version 1.5.3
-----
* Function ``pamer.fnc'' now accepts models with interactions
in the ranef structure (e.g., (1|X:Y).
* Added lower- and upper-bound degrees of freedom to ``posthoc.fnc''
results, as well as lower- and upper-bound unadjusted p-values
and lower- and upper-bound adjusted p-values (Bonferroni corrected).
* fixed a bug in function ``fitLMER.fnc'': the function does not set
the current working directory to ``tempdir()'' anymore and then leave
you there once the fitting is done.
--------------------
version 1.5.2
-----
* In function ``fitLMER.fnc'', changed argument ``file.name'' to
``log.file.name'' to make it more transparent.
* Disabled the automatic changing of the current working directory
to directory ``temp.dir'' in functions ``bfFixefLMER_F.fnc'',
``bfFixefLMER_t.fnc'', ``ffRanefLMER.fnc'', and ``fitLMER.fnc''.
* Argumants ``model'' and ``data'' now have no default value in
functions ``bfFixefLMER_F.fnc'', ``bfFixefLMER_t.fnc'',
``ffRanefLMER.fnc'', and ``fitLMER.fnc'' (before version 1.5.2,
these arguments had ``as.character()'' as a default).
* Column ``expl.dev.'' in function ``pamer.fnc'' now lists the
percentage explained (which is simply the value returned in version
1.5 times 100).
--------------------
version 1.5
-----
* Function ``pamer.fnc'' additioanlly returns the deviance explained,
R-squared, for each model term.
* Fixed a little bug with the fitting logs. The user can now change the
path to where the log will be saved.
* Added argument ``col = "red"'' to function ``mcp.fnc''.
--------------------
version 1.4
-----
* Removed dependency on package ``Design'', and consequently removed
any example that uses ``rcs'' or ``poly''.
--------------------
version 1.3
-----
* Added argument ``trim'' to functions ``mcp.fnc'' and ``romr.fnc'';
* Removed argument ``data'' in function ``mcp.fnc''. The data frame
on which the model was fitted is now automatically retrieved from the
model object itself (``model@frame'').
* Added values ``Upper Den. Df'', ``Lower Den. DF'', and ``Lower p value''
to function ``pamer.fnc''.
* Added argument ``llrt'' (log-likelihood ratio testing) to functions
``bfFixefLMER_F.fnc'', ``bfFixefLMER.fnc_t'', and ``fitLMER.fnc''.
* Added argument ``p.value'' to functions ``fitLMER.fnc'' and
``bfFixefLMER_F.fnc'' to allow the user to choose to backfit either
on upper-bound (anti-conservative) or lower-bound (consevative)
\emph{p''-values.
* Added function ``posthoc.fnc''.
* Added function ``perSubjectTrim.fnc''.