gfgmesevage - Network

Trying this site for #econometrics advice: I have a T=2 panel and I want to estimate the effect of a continuous treatment. The outcome has...

Trying this site for #econometrics advice:

I have a T=2 panel and I want to estimate the effect of a continuous treatment. The outcome has strong regression to the mean (cor(y2-y1,y1) = -0.5). In the simple binary treatment case the FE model in differences is

y2 - y1 = d*treat + e

and LDV is

y2 = b*y1 + d*treat + e, so if b != 0 the differenced model is misspecified. But with continuous treatment this is not as simple. Does regression to mean imply you should always use a LDV model?

My #introduction I’m a Lecturer (Assistant Professor) at King’s College London where I teach economic history. I work mostly on 19th...

My #introduction I’m a Lecturer (Assistant Professor) at King’s College London where I teach economic history. I work mostly on 19th century financial history in the UK. Currently working on market structure on the London Stock Exchange, the Poor Law, and inflation. I have unreasonable views on 19th century share price lists.

I like #rstats, #stan and Bayesian stats generally. And have worked on social networks in economic history. Desperately trying to disentangle homophily and contagion.