Adapted from “Vertical Relationships and Competition in Retail Gasoline Markets,” 2004 (Justine Hastings)ģ. Journal American Statistical Association.1980.Ģ. Randomization Analysis of Experimental Data in the Fisher Randomization Test. Perform sub-analysis to see if intervention had similar/different effect on components of the outcomeĮpi6 in-class presentation April 30, 2013ġ. Use robust standard errors to account for autocorrelation between pre/post in same individual Use linear probability model to help with interpretabilityīe sure to examine composition of population in treatment/intervention and control groups before and after intervention Requires baseline data & a non-intervention groupĬannnot use if intervention allocation determined by baseline outcomeĬannot use if comparison groups have different outcome trend (Abadie 2005 has proposed solution)Ĭannot use if composition of groups pre/post change are not stableīe sure outcome trend did not influence allocation of the treatment/interventionĪcquire more data points before and after to test parallel trend assumption (DID focuses on changerather than absolute levels)Īccounts for change/change due to factors other than intervention Violation of parallel trend assumption will lead to biased estimation of the causal effect.Ĭan obtain causal effect using observational data if assumptions are metĬan use either individual and group level dataĬomparison groups can start at different levels of the outcome. It has also been proposed that the smaller the time period tested, the more likely the assumption is to hold. Although there is no statistical test for this assumption, visual inspection is useful when you have observations over many time points. It requires that in the absence of treatment, the difference between the ‘treatment’ and ‘control’ group is constant over time. ![]() The parallel trend assumption is the most critical of the above the four assumptions to ensure internal validity of DID models and is the hardest to fulfill. ![]() Treatment/intervention and control groups have Parallel Trends in outcome (see below for details)Ĭomposition of intervention and comparison groups is stable for repeated cross-sectional design (part of SUTVA) Intervention unrelated to outcome at baseline (allocation of intervention was not determined by outcome) ![]() In order to estimate any causal effect, three assumptions must hold: exchangeability, positivity, and Stable Unit Treatment Value Assumption (SUTVA)1 Please refer to Lechner 2011 article for more details. The approach removes biases in post-intervention period comparisons between the treatment and control group that could be the result from permanent differences between those groups, as well as biases from comparisons over time in the treatment group that could be the result of trends due to other causes of the outcome.ĭID usually is used to estimate the treatment effect on the treated (causal effect in the exposed), although with stronger assumptions the technique can be used to estimate the Average Treatment Effect (ATE) or the causal effect in the population. DID requires data from pre-/post-intervention, such as cohort or panel data (individual level data over time) or repeated cross-sectional data (individual or group level). ![]() Hence, Difference-in-difference is a useful technique to use when randomization on the individual level is not possible. DID relies on a less strict exchangeability assumption, i.e., in absence of treatment, the unobserved differences between treatment and control groups arethe same overtime. Difference-in-Difference estimation, graphical explanationĭID is used in observational settings where exchangeability cannot be assumed between the treatment and control groups.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |