Skip to contents

Variance Partitioning Analysis

Source: R/varpart.R
varpart.Rd

Compute fraction of variance attributable to each variable in regression model. Also interpretable as the intra-class correlation after correcting for all other variables in the model.

Usage

varpart(
  fit,
  ...,
  distr.method = c("trigamma", "lognormal"),
  lambda.method = c("parametric", "mean")
)

# S4 method for class 'modelFitsResponses'
varpart(
  fit,
  fit_null,
  distr.method = c("trigamma", "lognormal"),
  lambda.method = c("parametric", "mean")
)

Arguments

fit

model fit

...

other args

distr.method

use either the "lognormal" or "trigamma" formulas from Nakagawa, et al. (2017)

lambda.method

use either "parametric" or "mean" method to estimate the mean rate for count models

fit_null

fit of null model

Details

The coefficient of determination (i.e. R^2) is 1 - [Residuals fraction]. This matches performance::r2_nakagawa() and performance::r2_mckelvey(), except these use the "lognormal" method.

For count models, the distributional variance is a function of the mean count rate. Following Eqn 5.8 of Nakagawa, et al. 2017, this can be estimated using parameters of a model including only intercept and random effect terms. But this requires refitting the model dropping the rest of the fixed effects. Instead, computing the mean of the observed counts is a fast approximation.

References

Nakagawa, Johnson, Schielzeth. 2017. The coefficient of determination R2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded. J. R. Soc. Interface 14: 20170213. doi:10.1098/rsif.2017.0213

Nakagawa, Shinichi, and Holger Schielzeth. "A general and simple method for obtaining R2 from generalized linear mixed‐effects models." Methods in ecology and evolution 4, no. 2 (2013): 133-142. doi:10.1111/j.2041-210x.2012.00261.x

See also

fastglmm::varpart()

Examples

library(fastglmm)
library(lme4)

set.seed(1)
sleepstudy$V <- rnorm(nrow(sleepstudy))

# prepare response, design and random effect
Y <- matrix(sleepstudy$Reaction, nrow=1)
rownames(Y) <- "Reaction"
colnames(Y) <- rownames(sleepstudy)

design <- model.matrix(~ Days, sleepstudy)
Z <- preprocess_indicator(sleepstudy$Subject)

fit <- glmmFitResponses(Y, design, Z, family=gaussian())

fit$design <- design
fit$formula <- ~ Days + (1|Subject)

varpart(fit)
#>               Days   Subject Residuals
#> Reaction 0.2845826 0.4055011 0.3099164