oguerrer / ppi

The model of Policy Priority Inference
MIT License
10 stars 5 forks source link

simple prospective simulation #9

Closed r-leyshon closed 1 year ago

r-leyshon commented 1 year ago

This may be my own misunderstanding, apologies if this is the case.

In tutorials/03 - simple_prospective.ipynb, there is the following cell:

sample_size = 100 # number of Monte Carlo simulations

outputs = []
for sample in range(sample_size):
    output = ppi.run_ppi(I0, alphas, alphas_prime, betas, A=A, R=R, qm=qm, rl=rl,
                Imax=Imax, Imin=Imin, Bs=Bs, B_dict=B_dict, T=T, G=goals)
    outputs.append(output)
oguerrer commented 1 year ago
  1. The cell that loads the data already loads I0 from the IF values of the indicators dataset.
  2. The tutorial on expenditure data explains that T is not the same as calendar time. In ehse turorial, in particular, because the model is calibrated with T=69, it means that every 3 computational steps equal one calendar year. Thus, to simulate 10 years in the future, T needs to be 30.
r-leyshon commented 1 year ago

@oguerrer many thanks for confirming. You are correct, the explanation of T is apparent within the expenditure tutorial. As you may have guessed, I skipped this due to lacking this data. This was a mistake as the guidance on the use of T was very useful.

I had missed the line on I0 = IF. I'll put this down to me reproducing your work with different data. Thanks for your help on this and sorry for all the issues.