Hello, I am attempting to the stteffects ipwra in order to study the effect of drug1 on time to death in a retrospective cohort with 10,000 patients (~5k per group).
stset deaddays_90, fail(dead_90==1)
stteffects ipwra ( i.age_cat preop_hbga1c i.insurance_type_simple i.smoking hx_chf hx_anemia i.hospital i.asacat3 dispo_simple i.proc_type_short healthyuser_pcp_last365_bin)(drug1 c.age##rx_insulin bmi female asa i.insurance_type_simple hx_pe hx_dvt hx_afib hx_copd hx_anemia hx_cad hx_oa hx_htn hx_cancer hx_asthma hx_gerd hx_depression hx_stroke hx_pvd hx_chf rx_betablocker rx_diuretic rx_ccb rx_statin rx_insulin preop_hbga1c rx_asa rx_coumadin rx_antiplt healthyuser_mamo_all healthyuser_pcp_last365_bin healthyuser_colonoscopy i.race) if fullcohort_rosey==1, vce(robust)
However, I believe I am having issues with convergence (I have been up to a thousand iterations w/o the EE criterion changing or getting an output).
Iteration 91: EE criterion = 1.368e-16 (not concave)
Iteration 92: EE criterion = 1.368e-16 (not concave)
Just looking at variables in logit or probit model, only requires 4 iterations. I have also tried taking out nearly all of the variables for bare-bones model testing and ma having the same problem.
stteffects ipwra (age preop_hbga1c )(drug1 age) if fullcohort_rosey ==1, vce(robust)
Any help or guidance would be greatly appreciated!!
Related Posts with iterations - stteffects ipwra
Significant AR(2) in System-GMM where the model does not include the lagged DVHi all, I am working on a system-GMM and the model does not include the lagged dependent variable. …
Stata Tabulate command & finding the most optimised modelHi statalists! I am very new to Stata - have just started my econometrics course and I am actually …
PPML reset testHello to all of you. I have been working in a gravity model with 133 countries from 2005 to 2018 (O…
PPML Reset Test ResultHello to all of you. I have been working in a gravity model with 133 countries from 2005 to 2018 (O…
Statistical formula for calculating 95% CIs from two geometric meansHello, I have the following data: Group 1: Mean = 8985, confidence intervals = 5705–14150 67 peopl…
Subscribe to:
Post Comments (Atom)
I am having the same issue with my stteffects models. I have tried running the same model as just a teffects and there are no identified overlap violators and results were produced immediately. Did you reach a solution to your problem?
ReplyDelete