The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
interplot()
now accepts a third variable via var3, plotting the
conditional effect of var1 across var2 at
several values (or levels) of var3. Fixing
var3 = c reduces the three-way effect to the existing
two-way problem with a shifted intercept and slope, so the same
simulation engine, stats_cp, and adjCI
machinery are reused. Continuous var3 defaults to the mean
and mean +/- 1 SD (Aiken and West 1991); factor var3 uses
all levels. facet = TRUE (default) draws faceted small
multiples; facet = FALSE overlays the curves colored by
var3. Supported for lm, glm,
mixed-effects (lmerMod, glmerMod),
multiply-imputed (lmmi, glmmi,
mlmmi, gmlmmi), and Bayesian
(brmsfit) models, with continuous
var1/var2.var1 interacts with a transformation of var2 –
a polynomial (var1:I(var2^2)), poly(), or a
spline (splines::ns(), bs()) –
interplot() now plots the conditional effect as the correct
nonlinear function of var2 instead of forcing a straight
line. The marginal effect at each moderator value is the model’s own
design-matrix gradient g(var2)' beta, evaluated from
posterior/simulation draws so that stats_cp and
adjCI continue to work. The existing hardcoded quadratic
(multiplier = 2) case is a special instance of this
gradient. Detected automatically; a plain two-way interaction is
unchanged. Because arm::sim() silently drops spline
(matrix-valued) coefficient columns, the nonlinear path samples
coefficients directly from N(coef, vcov)
(sim_betas_mvn). Supported for lm,
glm, mixed-effects, multiply-imputed (pooled across
imputations), and Bayesian (brmsfit) models.bin_layer() adds
the Hainmueller, Mummolo, and Xu (2019) binning estimator as a
composable ggplot2 overlay: within-bin (tercile) estimates of the
conditional effect of var1, drawn as dot-and-whiskers at
the bin medians. Departures from the interplot line flag
violations of the linear-interaction-effect assumption. Usage:
interplot(m, "var1", "var2") + bin_layer(m, "var1", "var2").jn_layer() now clips its non-significant shading to the
observed range of var2. Previously, when a Johnson-Neyman
bound fell far outside the data range, the shading rectangle stretched
the plot’s x-axis to that bound, leaving most of the panel empty.testthat (edition 3) suite covering three-way
interactions, the nonlinear gradient engine (including a 1e-12
equivalence guard that the new engine reproduces the existing linear and
quadratic outputs exactly), bin_layer(), the
jn_layer() axis clipping, and two-way backward
compatibility.brms::brm() via a new interplot.brmsfit()
method. The conditional coefficient is computed directly from the
model’s posterior draws (no arm::sim() re-simulation), so
the reported bounds are genuine equal-tailed posterior credible
intervals. Numeric and factor base terms, plot = FALSE,
stats_cp, hist, and predPro (via
brms::posterior_epred() for Bernoulli/binomial families)
are all supported. brms is a Suggests-only
dependency, so it is loaded lazily and is not required for the rest of
the package.jn_interval(). Identifies the values of the moderating
variable where the conditional effect crosses significance boundaries.
Works with lm, glm, lmerMod, and
glmerMod objects.jn_layer() for composable ggplot2 overlays:
vertical lines at JN bounds with shaded non-significant regions. Usage:
interplot(m, "var1", "var2") + jn_layer(jn).%o%, colMeans, apply) replacing
element-wise for loops. Shared helpers in utils.R
(sim_coef_vec, sim_diff_stats) eliminate
duplication across all method files.eval(parse(text = ...)) with direct
[[ indexing throughout for safety and clarity.summarize_all(funs(...)) with
summarize(across(...)).rbind loops with
purrr::map() + list_rbind().|>, lambda
\(x) syntax, inherits() for class
checking.abind and arm from
Depends to Imports so they are no longer
loaded and attached on library(interplot) (Issue #42).fake, coef1, ub,
lb) now gives a clear error message instead of a cryptic
failure (Issue #44).ci_diff and ks_diff undefined in
factor branches of interplot.lmmi,
interplot.glmmi, interplot.mlmmi, and
interplot.gmlmmi. These variables were referenced in
interplot.plot() calls but never computed in factor
interaction paths.test_cp not assigned in the histogram +
stats_cp = "ci" code path of
interplot.plot.size deprecation warnings: replaced
size with linewidth in geom_rect,
geom_errorbar, and geom_ribbon calls.predPro plots: the
x-axis sweeps var1, so it is now labeled with
var1 instead of var2.glm/lm dispatch order in
interplot(): glm is now checked before
lm since glm inherits from
lm.replicated_file_ks.R) that
would execute bare expressions on package load.In this version, the structure of the primary function is rewritten. The function treats factor and numeric base terms with separate side functions leaving room for more distinguished actions on the models with the two types of terms.
stats_cp = "ci" works for factor base-term models.plot = FALSE.predPro can be used on multilevel models.stats_cp = "ci" and stats_cp = "ks"
for lmerMod/glmerMod objects. The
ks_diff statistic was computed inside helper functions but
not returned, causing undefined-variable errors when
stats_cp was set to "ci" or "ks".
Now both ci_diff and ks_diff are properly
propagated from
extract_coef_numM/extract_coef_facM to
interplot.plot. The same fix is applied to the
lm/glm path for consistency.interplot.plot for the latest version of
ggplot2.globalVariables
declarations.stats_cp).txt_caption).set.seed() before calling interplot functions.class(m) == "polr".lmer projects.steps argument back in case of special design
requirement of the plot.Updated the vignette including instructions of how to change the aesthetics of the plot and how to use histogram function.
Updated the vignette including instructions of how to change the aesthetics of the plot and how to use histogram function.
geom_ functions.Adding the function to plot interactions based on factor variables.
## Bug fix Fit ggplot2 2.0.0 Fixed the quadratic error
(#16)
Adding the function to plot interactions based on factor variables.
## Bug fix Fit ggplot2 2.0.0
Fix the bug for nonlinear multilevel models with multiply imputed data (gmlmmi).
Fix the error to run mlm and mlmmi.
These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.