. 2022 Feb 12;31(4):411–423. doi: 10.1002/pds.5412

TABLE 4.

Comparison of the mean absolute standardised differences in the unweighted, pre‐defined, and pre‐defined and high‐dimensional propensity score (HDPS) weighted populations

Set of covariates	Accounting for relative importance of HDPS covariates ^a	Mean absolute standardised differences
Set of covariates	Accounting for relative importance of HDPS covariates ^a	Unweighted	Pre‐defined only weighted	Top 250 HDPS weighted	Top 500 HDPS weighted	Top 750 HDPS weighted
Pre‐defined only	–	7.74	0.11	1.56	1.51	1.68
Top 250 HDPS only	No	10.91	8.15	1.14	1.42	1.51
Top 250 HDPS only	Yes	6.73	5.11	0.62	0.77	0.88
Pre‐defined and top 250 HDPS	No	10.79	7.84	1.14	1.43	1.51
Pre‐defined and top 250 HDPS	Yes ^b	6.77	4.92	0.64	0.80	0.83

^{^a}

Given a ranked (e.g., Bross‐formula ranking) set of HDPS covariates of size N, importance weights are defined as ([N + 1] − rank)/N.

^{^b}

Predefined covariates are assigned an importance weight of 1.