Stata Command: testex
Stata Command: testex
Stata Command: testex.ado
A statistical test of the exclusion restriction of an instrumental variable (IV) based on D'Haultfoeuille, Hoderlein, & Sasaki (2021). Use it when you consider running an IV regression and want to test the exclusion restriction of the IV to ensure that it is a valid IV.
Installation:
. ssc install testex
Help:
. help testex
Reference: D'Haultfoeuille, X., Hoderlein, S & Sasaki, Y. (2021) Testing and Relaxing the Exclusion Restriction in the Control Function Approach. Journal of Econometrics, forthcoming. Paper
Title
testex -- Executes a statistical test of the exclusion restriction of an instrumental variable (IV).
Syntax
testex y x z [if] [in] [, numboot(real)]
Description
testex executes a statistical test of the exclusion restriction of an instrumental variable (IV) based on D'Haultfoeuille, Hoderlein, & Sasaki (2021). The command takes as input a scalar outcome variable y, a scalar endogenous variable x that is continuous, and a scalar instrumental variable z. If the instrument is not binary, then the program transforms the original instrument into a binary one. The output of the command includes the p-value of the test and the indicators of whether the test rejects the null hypothesis at various levels of statistical significance. The program gives a caution when there is no crossing of conditional CDFs required by Assumption 5 in D'Haultfoeuille, Hoderlein, & Sasaki (2021).
Option
numboot(real) sets the number of multiplier bootstrap iterations to compute the critical value of the test statistic. The default value is numboot(1000).
Examples
y outcome variable, x endogenous variable, z instrumental variable
Test of the exclusion restriction:
. testex y x z
Setting the number of multiplier bootstrap iterations to 2000
. testex y x z, numboot(2000)
Including additive controls:
. reg y controls . predict residy, resid . reg x controls . predict residx, resid . reg z controls . predict residz, resid . testex residy residx residz
Stored results
testex stores the following in r():
Scalars r(N) observations r(nb) Number of multiplier bootstrap iterations r(bw) Bandwidth r(KS) KS statistic r(p) P-value
Macros r(cmd) testex
Reference
D'Haultfoeuille, X., S. Hoderlein, & Y. Sasaki. 2021. Testing and Relaxing the Exclusion Restriction in the Control Function Approach. Journal of Econometrics, forthcoming. Link to Paper.
Authors
Xavier D'Haultfoeuille, CREST-ENSAE, Palaiseau, France. Stefan Hoderlein, Emory University, Atlanta, GA. Yuya Sasaki, Vanderbilt University, Nashville, TN.
testex -- Executes a statistical test of the exclusion restriction of an instrumental variable (IV).
Syntax
testex y x z [if] [in] [, numboot(real)]
Description
testex executes a statistical test of the exclusion restriction of an instrumental variable (IV) based on D'Haultfoeuille, Hoderlein, & Sasaki (2021). The command takes as input a scalar outcome variable y, a scalar endogenous variable x that is continuous, and a scalar instrumental variable z. If the instrument is not binary, then the program transforms the original instrument into a binary one. The output of the command includes the p-value of the test and the indicators of whether the test rejects the null hypothesis at various levels of statistical significance. The program gives a caution when there is no crossing of conditional CDFs required by Assumption 5 in D'Haultfoeuille, Hoderlein, & Sasaki (2021).
Option
numboot(real) sets the number of multiplier bootstrap iterations to compute the critical value of the test statistic. The default value is numboot(1000).
Examples
y outcome variable, x endogenous variable, z instrumental variable
Test of the exclusion restriction:
. testex y x z
Setting the number of multiplier bootstrap iterations to 2000
. testex y x z, numboot(2000)
Including additive controls:
. reg y controls . predict residy, resid . reg x controls . predict residx, resid . reg z controls . predict residz, resid . testex residy residx residz
Stored results
testex stores the following in r():
Scalars r(N) observations r(nb) Number of multiplier bootstrap iterations r(bw) Bandwidth r(KS) KS statistic r(p) P-value
Macros r(cmd) testex
Reference
D'Haultfoeuille, X., S. Hoderlein, & Y. Sasaki. 2021. Testing and Relaxing the Exclusion Restriction in the Control Function Approach. Journal of Econometrics, forthcoming. Link to Paper.
Authors
Xavier D'Haultfoeuille, CREST-ENSAE, Palaiseau, France. Stefan Hoderlein, Emory University, Atlanta, GA. Yuya Sasaki, Vanderbilt University, Nashville, TN.