Dynamic panel data - too many error messages

[helplessbunny]

Dear Wizards of Econometrics and EViews!

I am trying to estimate a LPM (since my dependent variable is binary) with a giant (~500000 obs. ~14000 crosssection ~80 periods) dataset and a lagged term of y with Arellano Bond I and Arellano Bond II - making it a DPD (hurray!)

This topic seems to be chewed through in this forum quite thoroughly, so if a solution already exists and I over-read I am sorry!

-) is it a problem that I want to estimate a binary dependent variable?
-) is it a problem that I include dummies in my regression?

-) the ONLY way I dont get an error message using ABI and ABII is when I ONLY use the lagged term in my regression - so what I put in the DPDWizard is:
.) no other regressors
.) one lag for y
.) no inclusion of period dummy variables (the box to tick)
.) no constant
.) differences
.) @dyn(y,-1,-2)
.) choose AB1 and before I estimate I delete y(-1) out of the equation box
if I dont do all of that I either get:
.) near singular matrix (I got that one SOOOOO may times)
.) order condition violated - insuffencient instruments (sometimes)
.) EViews computational time explodes
I dont know why I would get the first one - the last two I get probably because the number of instruments in relation to the N,T or # of obs. in total is too big

If anyone has an I idea what I could change or what I should tell the DPDWizard to do differently - I would greatly appreciate it and send you many virtual beers and teas!

Regards,
your humble reader

[helplessbunny]

for the next unlucky person that has to stumble across this post:

the problem might be that you have to tell, after all the wizard stuff is done, that the cross-section is not to be transformed (change from fixed to none in panel registry) - i guess this might be the cause for singularity in my case since i have a massive amount of cross sections