search

econ-ark / REMARK

Replications and Explorations Made using the ARK
Apache License 2.0
19 stars 56 forks source link

changing cstwMPC to new DistributionOfWealth repository #110

Closed sbenthall closed 3 years ago

sbenthall commented 3 years ago

This PR resubmits the old 'cstwMPC' repository, which was not up to the REMARK standard, with a new and updated repository, econ-ark/DistributionOfWealth.

The new repository uses HARK 0.11.0 and should be up to the new standard, with the exception that I have not yet tagged a release.

I request a review of this submission.

The main known issue is discrepancies between the numerical outputs of the original cstwMPC code and the new code. Original results: https://gist.github.com/sbenthall/ab205807836deed3dd30e1c914a6ddeb New results: https://gist.github.com/sbenthall/bcd40ea9d6eb6a4c67166d6d3941f024

@mnwhite says the small differences can be accounted for with changes to the stochastic simulation. But the somewhat larger differences between the MPC of the bottom end of the top 1/3 of the population suggested that perhaps there was a change to the accounting code that compiles and reports MPC distributions.

I was not able to find this difference.

If the new REMARK passes review, then I would:

sbenthall commented 3 years ago

I'll note that this REMARK can be improved by bringing in HARK's Calibration features to replace to old custom features. I would also like to one day refactor the execution code.

However, this repository as it is is compliant with the REMARK standard as currently stated, and would be a good starting point for that future work.

llorracc commented 3 years ago

Finally had a chance to look at this and agree with Matt that the differences are reassuringly small, easily within the reasonable bounds attributable to simulation noise. Am in agreement with the proposition that it would be good to update the antique hand-crafted calibration tools to use more modern tools now available in the toolkit, but we should first lay down this marker saying that just updating the code itself doesn't meaningfully change the results.

llorracc commented 3 years ago

@sbenthall, implicit in the prior message was an endorsement of your "I would" plans in your earlier message; this PS is to make the endorsement explicit.