Helper Functions for Subgroup Treatment Effect Pattern (STEP) Calculations

Usage

h_step_window(x, control = control_step())

h_step_trt_effect(data, model, variables, x)

h_step_survival_formula(variables, control = control_step())

h_step_survival_est(
  formula,
  data,
  variables,
  x,
  subset = rep(TRUE, nrow(data)),
  control = control_coxph()
)

h_step_rsp_formula(variables, control = c(control_step(), control_logistic()))

h_step_rsp_est(
  formula,
  data,
  variables,
  x,
  subset = rep(TRUE, nrow(data)),
  control = control_logistic()
)

Arguments

x

(numeric) biomarker value(s) to use (without NA).

control

(named list) from control_step().

data

(data.frame)
the dataset containing the variables to summarize.

model

the regression model object.

variables

(named list of string)
list of additional analysis variables.

formula

(formula)
the regression model formula.

subset

(logical)
subset vector.

Functions

• h_step_window(): creates the windows for STEP, based on the control settings provided. Returns a list containing the window-selection matrix sel and the interval information matrix interval.

• h_step_trt_effect(): calculates the estimated treatment effect estimate on the linear predictor scale and corresponding standard error from a STEP model fitted on data given variables specification, for a single biomarker value x. This works for both coxph and glm models, i.e. for calculating log hazard ratio or log odds ratio estimates. It returns a vector with elements est and se.

• h_step_survival_formula(): builds the model formula used in survival STEP calculations.

• h_step_survival_est(): estimates the model with formula built based on variables in data for a given subset and control parameters for the Cox regression, and returns a matrix of number of observations n, events as well as log hazard ratio estimates loghr, standard error se and Wald confidence interval bounds ci_lower and ci_upper. One row is included here for each biomarker value in x.

• h_step_rsp_formula(): builds the model formula used in response STEP calculations.

• h_step_rsp_est(): estimates the model with formula built based on variables in data for a given subset and control parameters for the logistic regression, and returns a matrix of number of observations n as well as log odds ratio estimates logor, standard error se and Wald confidence interval bounds ci_lower and ci_upper. One row is included here for each biomarker value in x.