pyrollmatch Extended Validation Report

Systematic comparison of pyrollmatch (Python) vs rollmatch (R) across 16 configurations.

Test	Config	Py Time	R Time	Py Matched	R Matched	Py max\|SMD\|	R max\|SMD\|	Score Corr	Pair Overlap
sample_size	n=100	0.018s	0.20s	98	96	0.1034	0.0970	0.999986	64.6%
sample_size	n=500	0.027s	0.48s	500	500	0.0422	0.0319	0.999932	17.5%
sample_size	n=1000	0.052s	1.02s	998	998	0.0261	0.0381	1.000000	61.4%
sample_size	n=2000	0.110s	2.25s	2000	2000	0.0189	0.0223	0.999991	15.1%
alpha	alpha=0.001	0.019s	0.38s	310	301	0.1004	0.0826	0.999932	3.1%
alpha	alpha=0.01	0.021s	0.39s	484	485	0.0496	0.0604	0.999932	14.5%
alpha	alpha=0.05	0.022s	0.41s	499	499	0.0442	0.0375	0.999932	17.1%
alpha	alpha=0.1	0.038s	0.47s	500	500	0.0422	0.0319	0.999932	17.5%
alpha	alpha=0.2	0.026s	0.50s	500	500	0.0420	0.0300	0.999932	17.5%
num_matches	matches=1	0.027s	0.38s	500	500	0.0422	0.0857	0.999932	7.2%
num_matches	matches=3	0.023s	0.46s	500	500	0.0422	0.0319	0.999932	17.5%
num_matches	matches=5	0.027s	0.54s	500	500	0.0357	0.0674	0.999932	24.3%
covariates	covs=3	0.028s	0.42s	498	498	0.0125	0.0464	1.000000	70.9%
covariates	covs=5	0.021s	0.47s	500	500	0.0422	0.0319	0.999932	17.5%
covariates	covs=10	0.024s	0.55s	496	494	0.0663	0.0468	0.999998	51.7%
no_replacement	repl=False	0.023s	0.36s	484	500	0.0524	0.0617	0.999932	7.4%

Interpretation

Score Correlation: Should be > 0.99. Measures agreement of propensity score models.
Pair Overlap: Varies by caliper tightness. Tight calipers → small differences in scores → different pairs.
max|SMD|: Both should be < 0.1 for valid matching. Agreement here matters more than pair identity.

Generated by pyrollmatch extended validation. Seed=42.