I noticed in the project roadmap that Linear IV Panel model estimation is listed as "not started." I would like to take this on as a substantial contribution.
My plan would be to start with the core estimators (FE-2SLS, RE-2SLS, pooled IV) and build incrementally from there, with tests validated against Stata and R. I am familiar with the existing panel and IV modules in the codebase.
Before I start, a few questions:
- Do you have any preferences on the API design? I assume it should follow the same patterns as
PanelOLS/IV2SLS (formula interface, .fit() returning a results object, etc.)
- Are there specific estimators you consider higher priority (e.g., Hausman-Taylor, Between-IV)?
- Should this live under
linearmodels/iv/ or a new linearmodels/panel/iv/ subpackage?
Happy to discuss the design before writing code.
I noticed in the project roadmap that Linear IV Panel model estimation is listed as "not started." I would like to take this on as a substantial contribution.
My plan would be to start with the core estimators (FE-2SLS, RE-2SLS, pooled IV) and build incrementally from there, with tests validated against Stata and R. I am familiar with the existing panel and IV modules in the codebase.
Before I start, a few questions:
PanelOLS/IV2SLS(formula interface,.fit()returning a results object, etc.)linearmodels/iv/or a newlinearmodels/panel/iv/subpackage?Happy to discuss the design before writing code.