search

mcaceresb / stata-honestdid

Robust inference in difference-in-differences and event study designs (Stata version of the R package of the same name)
70 stars 18 forks source link

Pull Request for #7: OSQP with defaults fails under some circumstances #8

Closed mcaceresb closed 1 year ago

mcaceresb commented 1 year ago

We uncovered in #7 that by default OSQP has a time-based switch for when to update a key parameter (documented here). OSQP's defaults are probably internally consistent, but we modified some of the default parameters to match CVXR (here); this includes a tighter tolerance for convergence.

As a result, this parameter is sometimes not updated in situations where the update is required for convergence (e.g. #7). The solution (with no adverse consequences that we could find) is to force the parameter to be updated every 25 iterations in the internal OSQP algorithm (25 is the current OSQP default if the time-based switch is activated early; see here and here).

@jonathandroth fyi

jonathandroth commented 1 year ago

Awesome - thanks!

On Mon, Nov 14, 2022, 2:18 PM Mauricio Caceres Bravo < @.***> wrote:

Merged #8 https://github.com/mcaceresb/stata-honestdid/pull/8 into main.

— Reply to this email directly, view it on GitHub https://github.com/mcaceresb/stata-honestdid/pull/8#event-7807355345, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE6EXFF22CQYGONSGVVR66LWIKF6TANCNFSM6AAAAAASAEXQG4 . You are receiving this because you were mentioned.Message ID: @.***>