brms (version 0.6.0)

hypothesis: Non-linear hypothesis testing

Description

Perform non-linear hypothesis testing of fixed effects parameters

Usage

hypothesis(x, hypothesis, class = "b", group = "", alpha = 0.05, ...)

Arguments

x

An R object typically of class brmsfit

hypothesis

A character vector specifying one or more non-linear hypothesis concerning parameters of the model

class

A string specifying the class of parameters being tested. Default is "b" for fixed effects. Other typical options are "sd" or "cor". If class = NULL, all parameters can be tested against each other, but have to be specified with their full

group

Name of a grouping factor to evaluate only random effects parameters related to this grouping factor. Ignored if class is not "sd" or "cor".

alpha

the alpha-level of the tests (default is 0.05)

...

Currently ignored

Value

• Summary statistics of the posterior distributions related to the hypotheses.

Details

Among others, hypothesis calculates an evidence ratio for each hypothesis. For a directed hypothesis, this is just the posterior probability under the hypothesis against its alternative. For an undirected (i.e. point) hypothesis the evidence ratio is a Bayes factor between the hypothesis and its alternative. In order to calculate this Bayes factor, all parameters related to the hypothesis must have proper priors and argument sample.priors of function brm must be set to TRUE. When interpreting Bayes factors, make sure that your priors are reasonable and carefully chosen, as the result will depend heavily on the priors. It particular, avoid using default priors.

Examples

fit_i <- brm(rating ~ treat + period + carry + (1+treat|subject),
             data = inhaler, family = "gaussian", sample.prior = TRUE,
             prior = set_prior("normal(0,2)", class = "b"), n.cluster = 2)

hypothesis(fit_i, "treat = period + carry")
hypothesis(fit_i, "exp(treat) - 3 = 0")

## perform one-sided hypothesis testing
hypothesis(fit_i, "period + carry - 3 < 0")

## compare random effects standard deviations
hypothesis(fit_i, "treat < Intercept", class = "sd", group  = "subject")

## test the amount of random intercept variance on all variance
h <- paste("sd_subject_Intercept^2 / (sd_subject_Intercept^2 +",
           "sd_subject_treat^2 + sigma_rating^2) = 0")
hypothesis(fit_i, h, class = NULL)

## test more than one hypothesis at once
hypothesis(fit_i, c("treat = period + carry", "exp(treat) - 3 = 0"))