Hi all,
I'm trying to define a Random within-between effect model (following Bell et al., 2019) on a 2 years-panel, where the time-variant predictors are demeaned within the Id over time, and the mean values are added as additional control to account for the between-variation. Therefore, the idea is to exploit both between and within variations in the workers over time.
However, starting with a simple random effect model, I'm having troubles when apply the pweights. Specifically, it takes too long to compute the standard errors and I have to stop the computation. Here the data and the model I'm actually fitting:
And I'm trying to run:
It converges, but keeps to compute standard errors forever, while it normally works if I re-run the same without weights. The same problem applies when I use the Random within-between effects.
What could be the reason for such failure? The weights are standard coefficients for representative population, where I use the first-entry year of individual weight.
I'm trying to define a Random within-between effect model (following Bell et al., 2019) on a 2 years-panel, where the time-variant predictors are demeaned within the Id over time, and the mean values are added as additional control to account for the between-variation. Therefore, the idea is to exploit both between and within variations in the workers over time.
However, starting with a simple random effect model, I'm having troubles when apply the pweights. Specifically, it takes too long to compute the standard errors and I have to stop the computation. Here the data and the model I'm actually fitting:
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input float(log_wage coeff_panel outsourced outfirm1 outworker1 ptime contract_type education) 7.600903 281.13644 0 0 0 0 4 3 7.863267 . 0 0 0 0 1 5 8.006368 . 0 0 1 0 4 2 7.17012 55.51101 1 1 1 0 4 2 7.377759 152.79167 0 0 1 0 4 5 7.090077 107.83526 0 0 0 0 4 6 7.423568 107.83526 0 0 0 0 4 6 7.377759 . 0 1 0 0 4 2 6.912743 . 0 0 0 1 4 2 7.582229 92.93236 0 0 1 0 4 2 end label values contract_type contract_type label def contract_type 1 "Permanent", modify label def contract_type 4 "Other temporary", modify label values education education label def education 2 "CEP Brevet des collèges, BEPC, CAP, BEP", modify label def education 3 "Bac tech ou profes. ou dipl.de ce niveau", modify label def education 5 "Bac+2", modify label def education 6 "Bac+3 ou Bac+4", modify
Code:
mixed log_wage i.gender outsourced outfirm1 outworker1 ptime i.contract_type i.education [pw=coeff_panel] || id:
What could be the reason for such failure? The weights are standard coefficients for representative population, where I use the first-entry year of individual weight.
Comment