construct_design_matrix_from_basis: Make a design matrix from a grouping factor and numeric time...

In wfforrest/maeve: Modeling Accessories Enabling Visualization of Experiments

Description

Takes a data frame with a grouping factor and a vector of time 'x'-values and a user choice of basis function type (currently 'tent' or 'poly') and returns a design matrix for fitting a piecewise linear mixed effects model. Tent basis vectors are created from the so-called "tent-functions". See, e.g., Simon Wood (2017) GAM 2nd edition text, section 4.2, page 164-166. Polynomial basis vectors (not to be confused with the gam spline approach) use the stats::poly() function. This is intended to be used in fitting linear mixed effects models to three (maybe four?) time points. It can in principal be used for more than that, but in that circumstance the spline approach will typically be better and is recommended.

Usage

construct_design_matrix_from_basis(
  study_data_frame,
  basis_choice = c("tent", "poly"),
  format_digits = 3,
  use_poly_coefs = FALSE,
  group_name = maeve_options("group_name"),
  x_name = maeve_options("x_name"),
  break_points = maeve_options("break_points"),
  number_break_points = maeve_options("number_break_points"),
  poly_degree = maeve_options("poly_degree")
)

Arguments

study_data_frame	data.frame with time ('x') vector and grouping vector
basis_choice	character describing which basis function type to use in making a design matrix.
format_digits	how many digits to retain in column names?
use_poly_coefs	logical whether to use "attr( maeve_options('poly_object'), 'coef')" to scale poly basis matrix.
group_name	character. See ?maeve_options().
x_name	character. See ?maeve_options().
break_points	numeric vector. See ?maeve_options().
number_break_points	numeric. See ?maeve_options().
poly_degree	numeric. See ?maeve_options().

Value

a design matrix, including intercept(s), that can be used with lme4::lmer().

Author(s)

Bill Forrest [e-mail]

See Also

matrix

Examples

   ### Make a basis for piecewise linear regression over 21 days from the vismodegib data:
      require( magrittr ) # want pipe operator "%>%"
      vismo_pwl_basis <-
        vismodegib %>%
        dplyr::filter( DAY_OF_STUDY <= 21 ) %>%
        maeve:::construct_design_matrix_from_basis( basis_choice = 'tent', 
                                                    break_points = c( 0, 7, 14, 21 )
                                                   )

wfforrest/maeve documentation built on Jan. 1, 2021, 12:47 p.m.