Code:
* Example generated by -dataex-. To install: ssc install dataex clear input float id byte(wave pre_post) float anyMMactivity_consistent byte(treatment_2 control_2) float(target close_rule) str24 b_tc_name 1 1 0 1 1 0 1 0 "Kibiito B" 1 2 1 1 1 0 0 0 "Kibiito B" 2 1 0 0 1 0 1 0 "Kibiito B" 2 2 1 1 1 0 0 0 "Kibiito B" 3 1 0 1 1 0 1 16 "Kibiito B" 3 2 1 1 1 0 0 16 "Kibiito B" 4 1 0 1 1 0 1 0 "Kibiito B" 4 2 1 1 1 0 0 0 "Kibiito B" 5 1 0 0 1 0 1 16 "Kibiito" 5 2 1 1 1 0 0 16 "Kibiito" 6 1 0 1 0 1 1 0 "Kibiito B" 6 2 1 1 0 1 0 0 "Kibiito B" 7 1 0 1 1 0 1 16 "Kibiito" 7 2 1 1 1 0 0 16 "Kibiito" 8 1 0 1 1 0 1 16 "Kibiito B" 8 2 1 1 1 0 0 16 "Kibiito B" 9 1 0 0 0 0 0 0 "Mujunju 2" 9 2 1 . 0 0 0 0 "Mujunju 2" 10 1 0 0 1 0 0 5 "Kibiito" 10 2 1 1 1 0 0 5 "Kibiito" end
Code:
xtivreg anyMMactivity_consistent i.pre_post##i.treatment_2 i.wave (pre_post#treatment_2 = close_rule) if treatment_2 == 1 | control_2 == 1, fe vce(cluster b_tc_name)
However, when I run the code from above I get the following error message: "depvars may not be interactions"
I also tried to do the estimation by hand:
Code:
gen interaction = pre_post*treatment_2 xtivreg anyMMactivity_consistent i.wave (interaction = close_rule) if treatment_2 == 1 | control_2 == 1, fe vce(cluster b_tc_name) Fixed-effects (within) IV regression Number of obs = 3,950 Group variable: id Number of groups = 1,975 R-sq: Obs per group: within = 0.3528 min = 2 between = . avg = 2.0 overall = 0.1952 max = 2 Wald chi2(1) = 915.11 corr(u_i, Xb) = 0.0000 Prob > chi2 = 0.0000 (Std. Err. adjusted for 108 clusters in b_tc_name) ------------------------------------------------------------------------------ | Robust anyMMactiv~t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- interaction | 0 (omitted) 2.wave | .4040506 .0133567 30.25 0.000 .3778719 .4302293 _cons | .5002532 .0066784 74.91 0.000 .4871638 .5133425 -------------+---------------------------------------------------------------- sigma_u | .30567061 sigma_e | .38718957 rho | .38395074 (fraction of variance due to u_i) ------------------------------------------------------------------------------ Instrumented: interaction Instruments: 2.wave close_rule ------------------------------------------------------------------------------
Code:
. ivreghdfe anyMMactivity_consistent i.pre_post##i.treatment_2 target (pre_post#treatment_2 = close_rule) if treatment_2 == 1 | control_2 == 1, absorb(i.id i.wave) cluster(b_tc_name) Warning - duplicate variables detected Duplicates: 1.pre_post#1.treatment_2 note: variable #2 is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09) note: variable #3 is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09) note: variable #6 is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09) (MWFE estimator converged in 2 iterations) OLS estimation -------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on b_tc_name Number of clusters (b_tc_name) = 108 Number of obs = 3950 F( 2, 107) = 2.87 Prob > F = 0.0611 Total (centered) SS = 295.7837975 Centered R2 = 0.0037 Total (uncentered) SS = 295.7837975 Uncentered R2 = 0.0037 Residual SS = 294.7025325 Root MSE = .2733 -------------------------------------------------------------------------------------- | Robust anyMMactivity_cons~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------------------+---------------------------------------------------------------- pre_post | 0 | 0 (empty) 1 | 0 (omitted) | treatment_2 | 0 | 0 (empty) 1 | 0 (omitted) | pre_post#treatment_2 | 0 0 | 0 (empty) 0 1 | 0 (empty) 1 0 | 0 (empty) 1 1 | .0709197 .030044 2.36 0.020 .0113611 .1304784 | target | .0171343 .025644 0.67 0.505 -.0337019 .0679705 -------------------------------------------------------------------------------------- Warning: estimated covariance matrix of moment conditions not of full rank. overidentification statistic not reported, and standard errors and model tests should be interpreted with caution. Possible causes: number of clusters insufficient to calculate robust covariance matrix singleton dummy variable (dummy with one 1 and N-1 0s or vice versa) partial option may address problem. ------------------------------------------------------------------------------ Collinearities detected among instruments: 3 instrument(s) dropped Included instruments: 1.pre_post 1.treatment_2 1.pre_post#1.treatment_2 target Excluded instruments: close_rule Partialled-out: _cons nb: total SS, model F and R2s are after partialling-out; any small-sample adjustments include partialled-out variables in regressor count K Duplicates: 1.pre_post#1.treatment_2 ------------------------------------------------------------------------------ Absorbed degrees of freedom: -----------------------------------------------------+ Absorbed FE | Categories - Redundant = Num. Coefs | -------------+---------------------------------------| id | 1975 1975 0 *| wave | 2 0 2 | -----------------------------------------------------+ * = FE nested within cluster; treated as redundant for DoF computation
I'd appreciate any help.
0 Response to Is it possible to instrument for a dummy/interaction in a DiD panel model with fixed effects?
Post a Comment