Predictions Using Nstool (NARX)

Hi, I am having trouble generating a prediction with the nstool app. I start by loading in two time series , setting my delay as 3 and then training the network. Supposedly with this, I should be able to predict future values of y(t). However with this tool i am only able to predict future values of y(t) if I actually already know what they are. For instance I train the network on a years worth of data for each time series. Then on the screen to perform additional tests with additional input data, I load in the previous ten days data for each time series up to today. When I then retest it only makes predictions for the ten days I have given it and not tomorrow. It seems it is only making predictions on values that it can evalute the error of its prediction on, however as I want tomorrows prediction today I do not have this. Technically it should be able to make this prediction as it says it can predict y(t) given values of y(t-1)....y(t-lag) and x(t-1)....x(t-lag) where t=tomorrow. Any help on how to get the program to output a single numeric prediction for tomorrow when I have no target value for it to compare its prediction to would be much appreciated.

Thanks in advance for your time to anyone who answers.

[Jake Plummer]

To anyone that sees this in the future and has the same problem, I think I may have a solution. Because the prediction of y(t) relys on y(t-1) and x(t-1) just pad each time series with one zero on the end. Run the program like normal and now the prediction for that day will be given. I have tried changing what number I pad with to see if it changes the outcome (it shouldnt) and it does not affect it so I am fairly sure it is ok to do. However if someone disagrees with this I would like to hear why?

Thanks again