LC References and Software Downloads
The LC approach is quite different from traditional Covariate Adjustment methods using Multivariable Models. In fact, rather than attempting to [1] fit parametric EQUATIONS to noisy health care data that fail to follow any sort of "law-like" relationships and to [2] predict overall MEAN values (treatment main-effects) from patient X-characteristics, the LC approach instead stresses flexible, non-parametric (robust), graphical analyses.
LC methods [a] estimate Local Treatment Differences (LTDs) only within X-space CLUSTERS of similar patients and [b] examine the joint DISTRIBUTION of these LTDs, thereby characterizing the full spectrum of patient differential response to treatment. Although somewhat computationally intensive, the LC approach is nevertheless ideal when databases are very large and encompass patients from numerous, diverse sub-populations. LC methods are based upon the Propensity Scoring theory of Rosenbaum and Rubin, Biometrika (1983). Specifically, cluster membership becomes a guaranteed BALANCING Score, "finer" than the propensity score, in the limit as clusters become small, compact and numerous!
The LC approach consists for FOUR PHASES of activities which (when repeatedly applied, checked and redone) ultimately assure that robust and objective (scientifically valid) results are being generated. The four links below display pages that introduce and illustrate the specific sorts of statistical analyses and graphical displays typically examined in each of the four phases of LC Analysis...
Phase One: Explore
Phase Two: Confirm
Phase Three: Agonize
Phase Four: Reveal
Want to learn more about methods for Observational Data Analysis? Purchase a copy of "the" book from SAS or Amazon!!!