Propensity scores derived in this fashion are called generalized propensity scores (GPS).
As the intensity of training measure is a continuous variable, we implement generalized propensity scores (GPS) with continuous treatments (Hirano and Imbens 2004) to estimate dose response functions on wage growth.
A propensity score weighted generalized linear model was used as the main multivariate analysis method; age, baseline A1c, copay, initial year, region, health plan, diabetes education, baseline comorbidities, diabetes medication, health care utilizations, and costs were controlled in the model.
Multivariate linear regression analyses with generalized estimating equations were performed after propensity score matching to balance covariates across classes of anchor agent.
To compare hospital and ward types, we used generalized linear mixed-effects models on a propensity score matched subset (n = 126,268) and on the total dataset.
After propensity score matching, covariance-adjusted and unadjusted generalized linear models (Dunnett-Hsu post-hoc analysis) were fitted to compare the mean values of laboratory parameters at baseline and during the exposure period in ARB users and CCB users, and were used to compare the mean change from the baseline value to the exposure value in ARB users and CCB users.
We calculated P values for differences between the two treatment groups after adjustment with the propensity score, including all 36 variables, estimated by generalized linear models (link id for continuous data and link logit for dichotomous data).
Stratified and regression analyses, including the propensity score to adjust for confounding, as well as generalized estimating equations to account for repeated vaccination, were used.
Further multiple logistic regression analyses with generalized estimating equation adjusted imbalanced baseline variables (p < 0.5 after matching) in propensity score matched patients [ 16] to confirm the head-to-head comparisons of early treatment discontinuation, treatment failure, and hematological AEs.
Propensity score weighting was used to adjust for respondents' propensity to be online.
