,
15 tweets,
14 min read
Read on Twitter
@EpiEllie @eschisterman1 @JohnPamplinII I was bummed too! I was really looking forward to their responses to my question/comment... 😏
@EpiEllie @eschisterman1 @JohnPamplinII I disagree with Miguel that most social epi exposures are analogous to the nutritional and biomarker examples he gave. For instance, his income example (this is supposed to be the easy one)...
@EpiEllie @eschisterman1 @JohnPamplinII His examples of hypothetical experimental manipulations leave out 3 considerations imp to #SocialEpi folks: 1) life course, 2) larger units of analysis (we are into health of communities, not just individuals), and 3) feedback mechanisms (if you give folks more $$,...3/
@EpiEllie @eschisterman1 @JohnPamplinII 3) feedback responses (I was going to quote a funny observation from Jay K in his Methods in Social Epidemiology book), if you give people money, their behaviors may change and even public policy may change (eg, Repub backlash to welfare expansions like the #ACA) 4/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan referred to a life course perspective later in his talk...5/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan And I think all these factors are compatible with a potential outcomes causal framework but commonly used methods in Epidemiology haven’t incorporated these well...6/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan Developing conceptual frameworks and methods for causal inference to better ask questions better incorporating life course, macro-level questions and feedback will serve social epi & the larger field. 7/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan Q: Do authors think current methods incorporate these complexities well? 8/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan Social epidemiologists are often interested in long term, large scale interventions that are not well-served by imagining hypothetical individual-level RCTs. 9/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan I personally think that qualitative, quasi-experimental, non-RCT thought experiments, etc are critical complements to traditional analysis of observational data for the kinds of questions we have #SER2019 10/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan Personally (bc human agency!) I don’t want to get *one* estimate of a causal effect on a racial health disparity ...12/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan I do the best studies I can w observational data, but I don’t trust it as a prediction of the hypothetical future...13/
@EpiEllie @eschisterman1 @JohnPamplinII @_MiguelHernan Instead I see estimates from my studies of race as contributing to a body of estimates for possible impacts of interventions, that has to be informed by outside knowledge from social theory, poli sci, behavioral theory, history, etc 14
