Talk:Propensity score

A nice extension to this article might be to include applications of the propensity score, e.g., to control for assignment bias. 20 years later, the propensity score is finally starting to take off in terms of use, but it is not a ubiquitous part of observational studies yet. --Andystats 23:07, 10 July 2007 (UTC)[reply]

'Unconfounded/unconfoundedness'[edit]

Neither of these are defined in the article, nor are there references to where they are defined. Rick lightburn (talk) 03:59, 13 October 2008 (UTC)[reply]

Practical interpretation[edit]

For those of us who are not trained in statistics, could an expert explain the following points? 1. What are the minimum and maximum values (if there are indeed limits) on the propensity score. Reply: It's a probability, so [0,1]. 2. What does a larger propensity score mean? Does it imply greater bias in assignment? Reply: the article only discusses a binary treatment, and propensity is probability of assignment to treatment 1, so higher score means more likely to be assigned to treatment 1. Both extremes of propensity would indicate bias; a value of .5 would mean no bias. —Preceding unsigned comment added by 63.125.124.226 (talk) 19:28, 7 November 2008 (UTC)[reply]