stepp.win: The constructor to create the stepp window object

Description

This is the constructor function to create a stepp window object.

Usage

stepp.win(type = "sliding", r1 = 5, r2 = 20, e1 = NULL, e2 = NULL)

Arguments

type: type of stepp window; "sliding", "sliding_events" or "tail-oriented"
r1: sliding window: minimum number of patients allowed in overlapping windows;
tail-oriented window: a vector of maximum covariate values for each subpopulation
r2: sliding window: size of subpopulation in each window;
tail-oriented window: a vector of minimum covariate values for each subpopulation
e1: sliding window: minimum number of events allowed in overlapping windows
e2: sliding window: number of events in each subpopulation

Author

Wai-ki Yip

Details

This is the functional interface to construct a stepp window object besides using new("stwin", ...). A STEPP window specifies a pattern of subpopulation you would like to explore. There are three kinds of patterns: "sliding window", "event-based sliding window" and "tail-oriented window".
These patterns are specified through the following set of values (r1 and r2 for sliding and tail-oriented windows, e1 and e2 for event-based sliding windows):

1. for sliding windows based on the number of patients ("sliding"), r1 is the minimum number of patients allowed in overlapping windows and r2 is the approximate size of subpopulation in each window;

2. for sliding windows based on the number of events ("sliding_events"), e1 is the minimum number of events allowed in overlapping windows and e2 is the approximate size of subpopulation in each window in terms of number of events;

3. for tail-oriented window, r1 is a vector of maximum covariate value for each subpopulation from the minimum value of the entire subpopulation, and r2 is a vector of minimum covariate values for each subpopulation from the maximum value of the window. The utility function, gen.tailwin, can be used to generate these two vectors based on the number of desired subpopulations.

When "sliding_events" is chosen r1 and r2 are set to NULL, while when "sliding" or "tail-oriented" are chosen e1 and e2 are set to NULL.

These pattern are based on all patients.

Examples

Run this code

  # create a stepp window object of type "sliding", 
  # subpopulation size is 200 and allows only 50 patients
  # between overlapping windows
  mywin <- stepp.win(type="sliding", r1=50, r2=200)

  # print a summary of the stepp window object
  summary(mywin) 

  # create a stepp window object of type "sliding_events", 
  # (event-based) subpopulation size is 200 and allows
  # only 50 events between overlapping windows
  mywin <- stepp.win(type="sliding_events", e1=50, e2=200)

  # print a summary of the stepp window object
  summary(mywin)