Backcasting MA errors with lagged variables

[rmoser]

I estimated the following model using MA backcasting turned on:

y c y(-1) x ma(1)

I now need to manually forecast y_hat values, and for that I need to compute the backcasted errors as done by eviews. I have read the eviews user manual and am able to do it in the absence of lagged variables, but I couldn't figure out what changes in this case.

¿Any ideas?

Thanks in advance!