Hi all. I have a panel dataset with socio-economic and political information on Brazilian municipalities (2002-2006-2010-2014-2018). I want to verify whether the number of higher education institutions in a municipality causes an increase in voting support for the left.
My outcome variable is “voto_pt_pres_perc” (vote for the Workers’ Party’s candidate for president) and my treatment variable is ”n_he_inst” (number of higher education institutions in a given municipality and year).
To that end, I want to estimate a dose response function with the command “doseresponse2”.
I have perused Guardabascio and Ventura’s (2013) paper “Estimating the dose-response function through the GLM approach” (https://mpra.ub.uni-muenchen.de/45013/), to no avail.
When I run the command, the estimation is not completed. The unsuccessful estimation follows below.
Can someone offer a help?
Thank you
********************************************
ESTIMATE OF THE GENERALIZED PROPENSITY SCORE
********************************************
Generalized Propensity Score
By agreement we assume that the logarithm of 0 is 0
************************************************** ****
Algorithm to estimate the generalized propensity score
************************************************** ****
Estimation of the propensity score
The log transformation of the treatment variable fraction is used
T
-------------------------------------------------------------
Percentiles Smallest
1% -5.062595 -5.062595
5% -5.062595 -5.062595
10% -3.676301 -5.062595 Obs 27,744
25% 0 -5.062595 Sum of wgt. 27,744
50% 0 Mean -.5273679
Largest Std. dev. 1.46311
75% 0 0
90% 0 0 Variance 2.140691
95% 0 0 Skewness -2.493764
99% 0 0 Kurtosis 7.40458
dependent variable T has negative values
r(499);
My outcome variable is “voto_pt_pres_perc” (vote for the Workers’ Party’s candidate for president) and my treatment variable is ”n_he_inst” (number of higher education institutions in a given municipality and year).
To that end, I want to estimate a dose response function with the command “doseresponse2”.
I have perused Guardabascio and Ventura’s (2013) paper “Estimating the dose-response function through the GLM approach” (https://mpra.ub.uni-muenchen.de/45013/), to no avail.
When I run the command, the estimation is not completed. The unsuccessful estimation follows below.
Can someone offer a help?
Thank you
Code:
qui egen max_p = max(n_he_inst)
gen fraction = n_he_inst/max_p
qui gen cut1 = 50/max_p if fraction<=50/max_p
qui replace cut1 = 100/max_p if fraction>50/max_p & fraction<=100/max_p
qui replace cut1 = 158/max_p if fraction>100/max_p
mat def tp1 = (0.10\0.20\0.30\0.40\0.50\0.60\0.70\0.80)
doseresponse2 coal_pres idhm pol_pi ln_r_pib_per_cap transfers pop_urb_perc ln_pop_tot schooling, outcome(voto_pt_pres_perc) t(fraction) gpscore(pscore) predict(hat_treat_nb) sigma(sd_nb) cutpoints(n_he_inst_cat) index(mean) nq_gps(5) t_transf(ln) dose_response(dose_response) tpoints(tp1) delta(1) reg_type_t(quadratic) family(binomial) link(logit) reg_type_gps(quadratic) interaction(1) bootstrap(yes) boot_reps(100) filename("output") analysis(yes) graph("graph_output") detail
********************************************
ESTIMATE OF THE GENERALIZED PROPENSITY SCORE
********************************************
Generalized Propensity Score
By agreement we assume that the logarithm of 0 is 0
************************************************** ****
Algorithm to estimate the generalized propensity score
************************************************** ****
Estimation of the propensity score
The log transformation of the treatment variable fraction is used
T
-------------------------------------------------------------
Percentiles Smallest
1% -5.062595 -5.062595
5% -5.062595 -5.062595
10% -3.676301 -5.062595 Obs 27,744
25% 0 -5.062595 Sum of wgt. 27,744
50% 0 Mean -.5273679
Largest Std. dev. 1.46311
75% 0 0
90% 0 0 Variance 2.140691
95% 0 0 Skewness -2.493764
99% 0 0 Kurtosis 7.40458
dependent variable T has negative values
r(499);
Comment