make_fake_data: Make fake dataset for testing and illustration

In li-xinran/RIQITE: Randomization Inference for Quantiles of Individual Treatment Effects

Make fake dataset for testing and illustration

Description

These are all fixed set-size and -outcome datasets. The pretend randomization is perfectly balanced, so we have two units of each type with one in treatment and one in control.

Usage

make_fake_data(scenario = "D")

Arguments

scenario

"A", "B", "C", or "D"; different kinds of data structure.

Details

The scenarios are:

A: Long right tail, but also some very low treatment units. Clear evidence of harm to some units and strong positive effects for others (the tx units are in the tails, co units are interior)

B: Treatment has lowest values, control has highest. It could be a simple shift down (constant negative tx effect visually plausible).

C: Treatment helps in general, and helps some a lot.

D: Treatment helps those at the low end, but hurts those at the high end. So testing for positive tx (alt = greater) on left tail should reject. Testing for negative tx on right tail(alt = less) should also reject.

Value

Dataset with Y0, Y1, Yobs, and Z (treatment assignment).

li-xinran/RIQITE documentation built on July 1, 2023, 6:58 p.m.