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.
eta2p_omnibus() computes a single
factor-level partial eta-squared for a multi-level factor or multi-df
interaction, corresponding to the omnibus (multi-df) test of that effect
rather than to individual contrasts. The variance attributed to the
effect is the variance of the summed fitted contribution of all of the
effect’s design-matrix columns, which correctly includes the covariances
among them; the error denominator matches eta2p(). Use this
when pairing an effect size with an omnibus F or chi-square test.eta2p() and batch_eta2p() now accept
factor predictors directly. The focal effect may be given either as a
model coefficient name (e.g. "Group75yr") or as the
underlying variable name (e.g. "Group"); a variable that
maps to a single coefficient is resolved automatically, and a
multi-level factor raises an informative error pointing to
eta2p_omnibus() or batch_eta2p(). Predictor
variances and factor random slopes are read from the model design matrix
and model frame, so manual recoding of factors to numeric is no longer
required.eta2p() gains an optional confidence interval via
parametric bootstrap (ci = TRUE), with
ci_level, n_boot, and seed
arguments. Because the effect size is a ratio of REML variance
components with no closed-form sampling distribution, the interval is
obtained by simulating from the fitted model
(lme4::bootMer), refitting, and recomputing the effect size
on each replicate. This is opt-in, as each replicate refits the
model.eta2p() and batch_eta2p() gain a
partial_predictors argument controlling the numerator for
single (non-interaction) predictors. When TRUE (the
default), the variance attributed to a predictor is its unique
(semipartial) variance – equivalently its total variance times its
tolerance, Var(X) * tol(X) – so it reflects only the
predictor’s unique contribution and declines with redundancy under
correlation. Set partial_predictors = FALSE for the
previous total-variance (raw) numerator. For centered, orthogonal
designs the two are identical, so this affects only models with
correlated predictors.partial_predictors above). For
centered, orthogonal designs this is identical to previous behavior; for
models with correlated predictors, the default effect sizes will be
smaller than in earlier versions, reflecting only each predictor’s
unique contribution. Set partial_predictors = FALSE to
reproduce the previous total-variance behavior.verbose in eta2p() is now
FALSE (previously TRUE), matching
batch_eta2p(). Pass verbose = TRUE to restore
printed output.operative = TRUE) now work
correctly for factor predictors. Within/between classification
previously matched the focal effect’s name directly against the data
columns, which failed for a factor whose coefficient name
(e.g. "Age75") differs from its variable name
(e.g. "Age"); the grouping factor could then be
misclassified and the wrong intercept variances included in or excluded
from the operative denominator. Coefficient names are now resolved back
to their underlying variable before classification.x1:x2) was silently omitted from the error
denominator when its random-effect name did not match the fixed-effect
product column. The contribution of interaction random slopes is now
resolved from the model frame and included.eta2p() and batch_eta2p() now support
design = "mixed" for models with both crossed and nested
random effects simultaneously. A canonical example is participants
viewing multiple photos of each model: photos are nested within models,
but both levels are crossed with participants. Supply both
cross_vars and nest_vars to use this
design.detect_within_between() previously used hardcoded
$subj and $item keys to classify grouping
factors as within or between, which caused intercept variances to be
silently omitted from the operative denominator. Keys are now indexed by
actual variable name, so the correct components are always
included.batch_eta2p() output columns for within/between status
are now named within_<varname> (e.g.,
within_participant, within_item) rather than
the hardcoded within_subj and within_item.
Code that references these columns by name will need to be updated.eta2p() computes partial eta-squared for a single fixed
effect in a fitted lmer model, supporting crossed and
nested random effects structures.batch_eta2p() computes partial eta-squared for all
fixed effects in a model.cross_vars.operative = TRUE.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.