Simplifying Assumptions Sample Clauses

Simplifying Assumptions. The stable-unit-treatment-value assumption described in section 1.1.4 is the first standard assumption necessary to move forward with causal estimates in this case as well. Two additional assumptions are not necessary to conduct causal estimates, but when applicable may narrow the estimated bounds. First is the monotonicity assumption (referred to as A1 in future chapters and sections), which states that there is no ‘denier’ or DL group - i.e., no patients would die under treatment but live if assigned to the control group. Under this assumption πDL = 0 so the new estimations of the remaining πs are: πLL = PcL πLD = PtL − PcL πDD = 1 − PtL This assumption may be conceptually appealing in clinical trails, since we obviously hope that a treatment does no harm, but it may or may not actually reflect the observed data. Second is referred to by Zhang and Xxxxx as ‘stochastic dominance’ or ‘ranked average score.’ [2003] This is the case that the group of always survivors (LL) is on average healthier than the rest of the sample, and therefore this subset should be considered to have at least as good of an outcome as the DL group under control and the LD group under treatment. Again, this assumption may or may not be applicable for any specific set of observed data (and is referred to as A2 in future chapters and sections).