Closed ijyliu closed 3 years ago
I'm sort of inclined to think that we shouldn't have a lit review at first. This might be worth asking Nadav or Bonhomme, but I haven't gotten the impression that a lit review is a requirement for this assignment?
@nicomarto do you have any more details from your conversation with the Prof. on how we would relate to https://warwick.ac.uk/fac/soc/economics/staff/knagasawa/PartialEffects.pdf ? I have gone through it (it's kind of tricky to read and I'm not really sure what partial effects are, maybe you know more). I don't really see how it connects to what we are doing- there is no mention of PCA or PCR. The only really related bits are that it uses a proxy variable and there's some error in covariates (nuisance parameters).
schennach- see heckman 2010a Here it is: seems very much like our situation: http://jenni.uchicago.edu/econ374/papers/Heckman_Schennach_Williams_2011_lin_lin_matching_2011-10-03a_cji.pdf
If possible, I'd like someone else to maybe check through this because it's not clear that I understood a lot.
I think @nicomarto is best suited for this, since he seems to know the most about it, and also because @marionoro has other stuff to work on.
yes, that jimmy's paper that Bonhomme told me about. I can take care of it definately, but first Id rather ask Nadav/Bonhomme how long lit rev should be considering that we are over 50% the page limit
I think the lit review in the introduction is fine in terms of length, but the lit review in the empirical section seems long.
Like, this paragraph is too much:
Guess I'll go cut it down now.
reopening this as I think @nicomarto was still looking at it?
I am happy with lit review and with what Paul wrote about the curse of dimensionality, so very good here
Even if we don't have to include literature review I think it can't possibly hurt us to have looked around and gained knowledge as to whether we're on the right track in terms of what we're doing.
Here are a lot of links (checking these off as i make a first pass skimming and taking brief notes):
Overall summary of findings: So it actually kind of seems like we are doing something unique, but it's not clear that what we are doing is better than existing approaches to measurement error. Like averaging or instrumenting variables with each other.