tx_function_factory | R Documentation |
The function returned is a function that returns a vector of
treatment impacts to go with passed vector of Y0 values. I.e.,
final function is of format: tx_function( Y0 )
.
tx_function_factory(distribution, ATE = 0, rho = 0, tx_scale = 1, ...)
distribution |
A string name of a distribution. "constant" is a constant treatment effect. "scale" is a scaling of Y0. "rexp" and "rnorm" are those random distributions, generated independent of the Y0. "linear" is a linear function of Y0; see above. |
ATE |
the average treatment effect. Ignored by "scale" option. |
rho |
Strength of relationship between the control-side distribution and the individual treatment impacts (assuming variation in both). NOTE: Depending on the details of how individual impacts are generated, rho may not be a strict correlation. See details of the impact generation functions for more. |
tx_scale |
Scaling factor for impacts. For rexp and rnorm, final treatment effects will have this standard deviation, in expectation. to achieve desired standard deviation of impacts generated by tx_function by (default is 1). Ignored by "constant" and "scale" options. |
... |
Extra parameters to pass to distribution function (e.g., sd for rnorm). |
For the linear option, the impacts are a linear function of Y0 of
tau = ATE + rho Y0 + epsilon,
with epsilon normally distributed with standard deviation of
tx_scale
.
For rexp and rnorm, impacts are generated with the passed distribution, and then ordered so if they were transformed to a normal distribution they would have about a rho correlation with the passed outcomes.
For scale, the impact is simply a scaling of the Y0 by "rho". This is equivalent to setting ATE = 0 and tx_scale = 0 with "linear", above).
For constant, all parameters other than the ATE are ignored.
A function of the form tx_function( Y0 ).
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.