get_indices: Helper function to return the correct indices for latent...

In awstringer1/casecrossover: Approximate Bayesian Inference for Case Crossover Models

Description

The organization of the latent field is confusing. There are random effects and fixed effects, plus the actual linear predictor values (which you usually don't want).

This function accepts a ccmodeldata object and returns a list telling you which indices in the vector of latent variables correspond to which model elements.

Details: the latent Gaussian variables have dimension Wd = Nd + M + p. Nd is the total number of control days, sum(model_data$control_days), of all subjects. M is the total number of random effects, i.e. unique values of smooth terms, in the order in which they appear in the model. p is the number of regression coefficients. To get the means/variances of the random effects, the correct call is i = (model_data$Nd+1):(model_data$Nd+model_data$M). To get the regression coefficients, the correct call is i = (model_data$Nd+model_data$M+1):(model_data$Nd+model_data$M+model_data$p).

All this is way too complicated for the user, and even for the developer, so here is a function to parse the model data and return an object which enumerates which elements of the (internal) latent field correspond to which quantities. This is mostly used internally.

Usage

get_indices(model_data, removezeros = TRUE)

Arguments

model_data	ccmodeldata object as returned by model_setup().
removezeros	Optional. Should the indices be returned with hard-zeroes included or removed from the random effect vector? Depends on where this function is being called.

Value

A list of class ccindex containing vectors of indices for the linear and smooth terms.

awstringer1/casecrossover documentation built on March 11, 2021, 4:41 a.m.