ld-archer
commented
1 year ago Problem with this that I didn't anticipate; models are estimated on the final_US data (including the education model), but we need the education model to predict max education on the final_US data. Initially I added a function to generate_stock_pop.py to do the education prediction for the stock pop, but I think I will create a separate script to do education prediction after creation of the replenishing population.
Currently max education prediction is only done for the replenishing population. This should also be done for young people in the stock population which have been overlooked.
This will be a function in minos.data_generation.generate_stock_pop that uses the education model to predict highest level of education for 16-25 year olds. Then some nonsense checking to remove anyone who is predicted a lower education than they currently have, and to account for those who have already passed an education threshold (e.g. A levels in the model are achieved by age 19, so a 23 year olds without A-levels should not be assigned A-levels. I know this could happen in real life, but we have made some assumptions for simplicity's sake).