[Mediation-information] Causal mediation analysis?

[Kosuke Imai]

Ho-Chuan,

  It is only applicable in the sense that conditional on pre-treatment covariates the treatment assignment must be ignorable.  If that condition is not met, then causal mediation analysis will be biased.   So, the key is to collect confounding variables so that the treatment can be assumed to be as-if random given those variables.

Kosuke

Department of Politics
Princeton University
http://imai.princeton.edu


On Oct 23, 2011, at 11:05 PM, River wrote:

> Dear Professor Imai,
>  
> I find that many of your papers related to causal mediation analysis are interesting and helpful. However, I wonder if the approach is applicable to the case where the assignment to treatment and control is nonrandom. For example, suppose that I am interested in measuring the macro-economic (treatment) effects (and the corresponding causal mechanisms) of inflation targeting (IT). However, it is well known that the choice of implementing the IT framework is self-selected by the central bankers and unlikely to be random.
>  
> Best wishes,
>  
> Ho-Chuan (River) Huang
> Professor, Department of Banking and Finance, Tamkang University
> 151, Yingzhuan Rd., Danshui Dist., New Taipei City 25137, Taiwan, R.O.C.
> <16_f_phone.gif>26215656 ext. 3337 <08_h_house.gif>http://mail.tku.edu.tw/river/
>  
>  
> <image.gif>
> <05_puppy_sticky_note_en.gif>