v0.2.2

OptimalGIV v0.2.2

Diff since v0.2.1

New Features

🎯 Principal Component Extraction Support (#9)

• Added support for internal principal component extraction from
  residuals using HeteroPCA.jl
• Use pc(k) in formulas to extract k common factors: @formula(q +
  endog(p) ~ controls + pc(2))
• PC extraction changes moment conditions from E[u_i u_{S,-i}] = 0 to
  E[u_i u_j] = ΛΛ'
• Includes PC factors, loadings, and HeteroPCA model in results
• Configure PC extraction with pca_option parameter

📐 Improved Architecture (#8)

• Refactored ObservationIndex into standalone utils module for better
  code organization
• Enhanced panel data handling and indexing capabilities

Bug Fixes

• Fixed bugs in joining fixed effects DataFrame (fedf) with main results
• Improved handling of missing values in PC extraction (missing values
  remain missing rather than zero)
• Fixed consistency issues between different estimation algorithms

Enhancements

• Consolidated and streamlined interface tests for better coverage
• Added comprehensive tests for PC extraction functionality
• Improved simulation framework with PC extraction support
• Enhanced documentation with CLAUDE.md for AI-assisted development
• Better error handling and validation for complex formulas

Breaking Changes

• Variance-covariance matrix calculation is automatically disabled when
  using PC extraction

Notes

• Small sample PC extraction may not have exact roots - consider using
  minimization instead
• Models with fully flexible elasticities and factor loadings are not
  theoretically identifiable

Merged pull requests:

• Refactor ObservationIndex into standalone utils module (#8) (@FuZhiyu)
• Add Internal Principal Component Extraction Support (#9) (@FuZhiyu)
• consolidate tests (#10) (@FuZhiyu)
• fix bugs in joining fedf and add more tests for saving (#11) (@FuZhiyu)