The first step of TPI

[JLian401]

From chapter 8.6, we know that we should guess BQ in the first step of TPI. Currently, we have a get_BQ function in aggregate.py, where we input r and other parameters to return BQ. But I think that is not we can use in the first step of TPI and that we should define it in the SS function. Is it the right way?

[jdebacker]

@JLian401 Yes, at least for the time path solution, you will need to guess the path of BQ. You can start with an initial guess of, e.g., BQ_path = np.ones(T) * BQ_ss, but then with each iteration through the time path you will use the get_BQ function to update this guess, just as you did with the r_path.

[saharnazb]

I started like this: (But not sure if I am doing right with (T + S)

[jdebacker]

@saharnazb That look like a good starts. But you'll want both the r_path and BQ_path to be the same length, T+S.

[JLian401]

@saharnazb In my understanding, maybe we can just set BQ_init as one input then make BQ_path = np.ones(T+S) * BQ_ss. Then we will no need to define BQ_path_init elsewhere and no need to use np.append .

[saharnazb]

@JLian401 That can make it simple. let's make sure it does not make any conflict with anything else.

[saharnazb]

@JLian401 I am thinking that we should keep _path_init because using _init the convergence may take longer but _path_init is defined based on r_ss. To get r_path, starting from _path_init will make the convergence faster

[saharnazb]

@jdebacker I am trying to use np.tri() for the time path solution. Is this the right way to solve? I was thinking to have zeros at and below the first full lifetime decision-maker. Or should I have the partial decision-makers' path separately?

[jdebacker]

@saharnazb You won't need to use np.tri(). Look a the 3 period solution , use a similar "masking" matrix...

[saharnazb]

@jdebacker I mean for the ones that are not full lifetime. I already defined them using np.tri() and iterated on t. So I should change it. But it seems someone else has done it in a different way. Should I continue?