Skip to contents

Methods for fitting “Sum of Single Effect” (SuSiE) models and computing posterior statistics from SuSiE models. The ‘fit_susie_model’ method is the main method for fitting SuSiE models and computing posterior statistics.

Usage

init_susie_model(
  variable.names,
  L = 10,
  ser = init_ser_model(variable.names, family = "normal", control = list(maxiter = 1,
    trace = FALSE)),
  residual_variance = 1,
  control = list()
)

fit_susie_model(fit, dat, return_data = missing(dat))

susie_control_default()

Arguments

variable.names

The names of the variables in the SuSiE model.

L

An integer, 1 or greater, specifying the number of single effects.

ser

An object of class ‘ser’ or a list of ‘ser’ objects specifying the SER family for each of the L single effect regressions. If ser is a list, it should have L elements.

residual_variance

Initial estimate of the residual variance parameter.

control

An object, possibly a list, specifying settings that affect the model fitting. See ‘Details’ for more information.

fit

An object of class ‘susie’ that captures the state of a SuSiE model fit, typically the output from running an ‘init_susie_model’ function, or an output from a previous fit_susie_model call.

dat

An object of class ‘mlm_data’, for example the output from a call to prepare_mlm_full_data or prepare_mlm_summary_data. See mlm_data for other ways to create ‘mlm_data’ objects.

return_data

If return_data = TRUE, the input data are included in the output.

Value

init_susie_model and fit_susie_model return objects of class ‘susie’ storing key posterior statistics for all variables and the state of the SuSiE model fit. It is a list with the following elements:

ser

The L single effect regression models which define the SuSiE posterior distribution.

residual_variance

The residual variance parameter.

elbo

The value of the ELBO (an approximate marginal likelihood) at the current model fit.

lbf

Approximate log-Bayes factor comparing the SuSiE model to the “null” model (b = 0).

progress

A data frame containing detailed information about progress on the model fitting.

control

A list containing the settings that affect how the model fitting is performed.

version

The version of the ebverse R package that was used to generate this object

dat

When return_data = TRUE, this contains the data that were used to fit the model.

Details

TO DO: Explain that the special case of L = 1 reduces to a single effect regression.

Examples

# TO DO: Add one or several examples.

# TO DO: Add example(s) here.