ivse: Functions to calculate the S.E. for IV estimates

In sumtxt/causaltools: Helpful functions for causal inference

Description

ivse calculates the S.E. for IV estimates based on estimates of the intention-to-treat effects. Uses the formula from Rubin/Imbens (2017, p. 531).

Usage

ivse(itty, ittw, ittyse, ittwse, ittcov)

ivse_grid(itty, ittw, ittyse, ittwse, df = TRUE, n = 100)

ivse_max(itty, ittw, ittyse, ittwse)

ivse_min(itty, ittw, ittyse, ittwse)

Arguments

itty	the ITT on the outcome
ittw	the ITT on the instrument (also known as complier share)
ittyse	the standard error of the ITT on the outcome
ittwse	the standard error of the ITT on the instrument
ittcov	the covariance between the intention to treat effects.
df	return data.frame (default) or vector of values?
n	number of covariance values within bounds

Details

The covariance between the intention-to-treat effects is bounded by the Cauchy–Schwarz inequality. ivse_grid calculates the S.E. for a sequences of covariances within these bounds.

ivse_min and ivse_max return the maximum/minimum S.E. when the covariance between intention-to-treat effects is unknown.

These functions are useful for two-sample two-stage least squares (2S2SLS) problems.

Value

numeric double vector or data.frame.

Functions

• ivse_grid:

• ivse_max:

• ivse_min:

Examples

 ## Not run: 

	Example from Imbens/Rubin, Chapter 23 
 (Vitamin A Supplemental Data)

 ittw = 0.8 
 ittwse = 0.0036 
 
 itty = 0.0026
 ittyse = 0.0009
 
 ittcov = -0.00000017

 ivse(itty,ittw,ittyse,ittwse,ittcov)


 
## End(Not run)

sumtxt/causaltools documentation built on May 14, 2019, 8:02 a.m.