QuantEcon / QuantEcon.py / 28091698120

93%

push

github

Add tests for Kalman.stationary_coefficients (#837)

* Add tests for Kalman.stationary_coefficients

Closes #128 by covering MA/VAR coefficient recursion, both Riccati
solvers, and invalid coeff_type handling.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Address review feedback on Kalman coefficient tests

Use closed-form matrix_power references instead of duplicating the
production recursion, and switch invalid coeff_type checks to
pytest.raises.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Add asymmetric Kalman coefficient test from lecture example

Cover stationary_coefficients with the pooling-equilibrium state space
from "Knowing the Forecasts of Others", alongside the symmetric toy
model.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Increase lecture-model H for scipy inv stability on CI

Use 1e-4 instead of 1e-8 to avoid singular-matrix errors from
scipy.linalg.inv on Python 3.14 in the conda CI environment.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Use two-noisy-signals filter for asymmetric Kalman test

Replace the pooling-equilibrium state space with the simpler (n, k) =
(1, 2) model from lecture section 37.5.3, using natural observation
noise H = np.eye(2) * sigma_e.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Address review feedback on two-noisy-signals Kalman tests

Set C so CC' matches Var(v_t), and compare stationary_coefficients
against lecture-derived K and matrix-power references.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Solve lecture posterior variance p with explicit quadratic formula

Replace solve_discrete_riccati in the two-noisy-signals reference with
the closed-form root from lecture eq. (37.26).

Co-authored-by: Cursor <cursoragent@cursor.com>

* Remove redundant non-diagonal coefficient sanity check

Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Co-authored-by: Cursor <cursoragent@cursor.com>

95 of 96 new or added lines in 1 file covered. (98.96%)

7994 of 8595 relevant lines covered (93.01%)

2.79 hits per line