Hi guys, I am fitting an ordered probit model and have discovered some specific variables meet parallel lines assumption by using the autofitting feature of regoprob2. However when I try to estimate the partial proportional odds model (using either autofit, autofit2 or by specifying variables using regoprob2) convergence is not achieved:
. regoprob2 ghql nkids female rfihhyr age agesq mastatd2 mastatd3 mastatd4 mastat
> d5 mastatd6 neduc poeduc oeduc aeduc oheduc reg1x reg2x reg3x reg4x reg5x reg6x
> reg7x reg8x reg9x reg10x reg12x, i(pid) pl(mastatd3 nkids age mastatd4)
convergence not achieved
convergence not achieved
r(430);
Any ideas how I could troubleshoot this problem or fit a generalised RE ordered probit model where the parallel lines assumption only applies to specific variables another way?
. regoprob2 ghql nkids female rfihhyr age agesq mastatd2 mastatd3 mastatd4 mastat
> d5 mastatd6 neduc poeduc oeduc aeduc oheduc reg1x reg2x reg3x reg4x reg5x reg6x
> reg7x reg8x reg9x reg10x reg12x, i(pid) pl(mastatd3 nkids age mastatd4)
convergence not achieved
convergence not achieved
r(430);
Any ideas how I could troubleshoot this problem or fit a generalised RE ordered probit model where the parallel lines assumption only applies to specific variables another way?
