Hello -- I am running an melogit on a 330,000 person nested in ~1900 neighborhoods in 47 countries. For my models, I keep getting "Initial Values Not Feasible". Below are the recommendations I have tried from other forums. I also randomly selected 50% of my sample and it ran then. However, this is not a really feasible solution. Any help is greatly appreciated !!!

Code:
melogit YNipv urban femalemore malemore working Zsurveyyr [pw=dvwgt] || country: || newid: , or nolog

logit YNipv zfemeduc zage working Zsurveyyr childtot urban [pw=dvwgt]
mat a=e(b)
mat a1=(a,0)
melogit YNipv zfemeduc zage working Zsurveyyr childtot urban [pw=dvwgt] || country: ///
|| newid: , or from (B) intmethod(laplace)

melogit YNipv zfemeduc zage working Zsurveyyr childtot urban || country: || newid: ,noestimate
matrix define B = e(b)
matrix define B[1,4] = 1e-8
matrix b1=e(b)
melogit YNipv zfemeduc zage working Zsurveyyr childtot urban [pw=dvwgt] || country: ///
|| newid: , or from (B, skip)

melogit YNipvemo zfemeduc zage working Zsurveyyr childtot urban [pw=dvwgt] || country: ///
|| newid: , or startgrid(2)