greta.glmer: Initialize the linear predictor for generalized linear effect...
In dirmeier/greta.lme: Linear mixed models in greta

Description Usage Arguments Value Examples

greta.glmer creates the linear predictor $eta$ required for fitting generalized linear mixed models. The linear predictor has the form:

η = X β + Z γ,

where X is the fixed effects design matrix, β the fixed effects, Z the random effects design matrix and γ the random effects.

1 2	greta.glmer(formula, data, prior_intercept = NULL, prior_coefficients = NULL, prior_random_effects = NULL)

`formula`	an `lme4`-style formula
`data`	the data set containing the variables described in `formula`
`prior_intercept`	a `greta_array` random variable that specifies the prior for the intercept β_0 for the fixed effects design matrix. If `NULL`, the distribution for the intercept is chosen as p(β) ∝ const. The intercept prior needs to have dimension 1.
`prior_coefficients`	a `greta_array` random variable that specifies the prior for the coefficientsβ for the fixed effects design matrix. If `NULL`, the distribution for the coefficients is chosen as p(β_i) ~ N(0, σ), and p(σ) ~ Half-Cauchy(0, ∞). The coefficients prior need to have the same dimensionalty as the number of columns of your fixed effects design matrix (without the intercept).
`prior_random_effects`	a `greta_array` random variable that specifies the prior for the random effects γ for the random effects design matrix. If `NULL`, the distribution for the random effects is chosen as p(γ) ~ N(0, τ), and p(τ) ~ Inv-Wishart If `prior_random_effects` is provided it needs to have the same dimensionalty as the number of columns of Z random effect design matrix. These are generally not trivial to construct. Calling `glFormula(formula, data)` creates the specified random effects matrix Z. From this the dimensionality of `prior_random_effects` can be inferred.

Returns a list of class greta.glmer with the following elements:

`predictor`	the linear predictor η
`X`	the fixed effects design matrix
`x.eta`	the linear predictor Xβ
`coef`	the prior for the coefficients β
`Z`	the random effects design matrix
`z.eta`	the linear predictor Zγ
`gamma`	the prior for the random effects γ
`Ztlist`	a list of random effect terms used for initializing γ
`grp.vars`	the variables used for grouping
`call`	the function call used
`formula`	the formula used

# create a random intercept model
greta.glmer(Sepal.Length ~ Sepal.Width + (1 | Species), iris)

# creates a random slope model
greta.glmer(Sepal.Length ~ Sepal.Width + (Sepal.Width | Species), iris)

# creates a random slope model with multiple random effect terms
greta.glmer(Sepal.Length ~ Sepal.Width + (Sepal.Width | Species) + (Petal.Width | Species),
            iris)

# creates a random slope model with strong regularizing prior
greta.glmer(Sepal.Length ~ Sepal.Width + (Sepal.Width | Species),
           iris,
           prior_random_effects = greta::normal(0, 1, dim=6))

# creates a random slope model with flat coefficient prior
greta.glmer(Sepal.Length ~ Sepal.Width + (Sepal.Width | Species),
           iris,
           prior_coefficients = greta::variable(dim=1))

# creates a random slope model with normal intercept prior
greta.glmer(Sepal.Length ~ Sepal.Width + (Sepal.Width | Species),
           iris,
           prior_intercept = greta::normal(0, 1))