clique_diagnostic: Diagnostic for Randomized Experiment

In dpuelz/CliqueRT: Randomization-Based Testing of Causal Effects under General Interference

Diagnostic for Randomized Experiment

Description

For different number of units selected, it returns an fmax by 2 matrix, where fmax is the maximum number of units to be selected (see below). The first column is the approximate number of focal assignments of the largest possible biclique of the null exposure graph that takes the selected units as focal units if we draw N randomizations. The second column is the first column with an additional requirement that each of the biclique's focal assignments does not contain only one type of exposure. It can serve as a diagnostic for choosing minr, minc or minass in the function clique_test.

Usage

clique_diagnostic(
  struc,
  p_group,
  p_indi_t,
  p_indi_nt,
  N,
  fmax = 25,
  NR = 500,
  Nx = 500
)

Arguments

struc	Structure of the cluster. It should have two columns. The first column represents group ID, and the second column is the individual ID. Please see the example.
p_group	The probability of a group being selected as treated group.
p_indi_t	The probability of an individual in the treated group being treated.
p_indi_nt	The probability of an individual in the non-treated group being treated.
N	The number of randomizations performed.
fmax	Maximum number of units to be selected to do the diagnosis. Default is 25.

Value

an fmax by 2 matrix.

dpuelz/CliqueRT documentation built on Jan. 6, 2023, 11:20 p.m.