nauf-package: Regression with NA values in unordered factors.

In CDEager/nauf: Regression with NA Values in Unordered Factors

Description

It is often the case that a factor only makes sense in a subset of a dataset (i.e. for some observations a factor may simply not be meaningful), or that with observational datasets there are no observations in some levels of an interaction term. There are also cases where a random effects grouping factor is only applicable in a subset the data, and it is desireable to model the noise introduced by the repeated measures on the group members within the subset of the data where the repeated measures exist. The nauf package allows unordered factors and random effects grouping factors to be coded as NA in the subsets of the data where they are not applicable or otherwise not contrastive. It is highly recommended that variables be put on the same scale with standardize prior to using nauf functions (though this is not required).

Contrasts

A detailed description of how NA values are treated is given in nauf_contrasts. These contrasts are implemented automatically through nauf_model.frame, which stores the information required to make fixed effects and random effects model matrices with nauf contrasts in its terms attribute. For details on the terms attribute, see nauf.terms. For fixed effects and random effects model matrices, see nauf_model.matrix and nauf_glFormula.

Regressions

nauf contrasts have been implemented for fixed effects regressions that would normally be fit with lm, glm, and glm.nb; and for mixed effects regressions that would normally be fit with lmer, glmer, and glmer.nb. For fixed effects nauf regressions, see nauf_glm, and for mixed effects nauf regressions, see nauf_glmer. There is also support for Bayesian versions of these models as would normally be fit with stan_lm, stan_glm, stan_glm.nb, stan_lmer, stan_glmer, and stan_glmer.nb (see nauf_stan_glm and nauf_stan_glmer for details).

ANOVAs

The anova function can be used with any frequentist (i.e. not Bayesian) nauf model. For fixed effects nauf models, the anova function uses the methods for corresponding non-nauf models (i.e. anova.lm, anova.glm, or anova.negbin depending on the regression family). For mixed effects nauf models, the anova function uses the anova.nauf.merMod method, which with default arguments is the same as anova.merMod. The anova.nauf.merMod method also allows Type III tests to be made via likelihood ratio tests, parametric bootstrapping, and, for linear models, the Satterthwaite and Kenward-Roger approximations of denominator degrees of freedom (similar to the mixed function in the afex package).

Predicted Marginal Means

The nauf_ref.grid function can be used to construct reference grids for nauf models (constructing a ref.grid-class object). Predicted marginal means (often called least-squares means; but in the case of Bayesian regression they are posterior marginal means) can be calculated with this reference grid using the nauf_pmmeans function. The function also allows the user to flexibly specify a subset of the reference grid to use when calculating the marginal means, so that the effect of a factor can be tested with regards to the subset of the data where it is contrastive.

Datasets and Vignette

For detailed examples of how to use the nauf package, see the "Using the nauf package" vignette, which makes use of the plosives and fricatives datasets included in the package.

CDEager/nauf documentation built on May 6, 2019, 9:24 a.m.