Write Up Empirical Results

[paul-opheim]

Isaac, could you run the regression results with covariate options that match those from the simulation section (so four options: a single mis-measured covariate, PCA on all covariates, the average of all mis-measured covariates, including all mis-measured covariates together)?

Then, could you include the results from those regressions in a table in the paper, and write up commentary around those results in the style of writing that is consistent with what we want our final paper to look like? I think this would basically involve fleshing out the bullet points that you've already written, along with having a brief discussion of the regression results after you show the table. I think it would be helpful to have the introduction, theory, simulation, and empirical results all written up so that we can see what the paper would look like.

• [x] Minor edits to the regression table abbreviations- GDP PC and PCA as covariates names. Possibly use same names as Paul.

[ijyliu]

I'm kind of unsure on how to proceed with 'the average of all mis-measured covariates'. Some of our covariates are GDP, the others are poverty levels and foreign aid levels, etc. and it wouldn't really make sense to average those.

Unless you mean either:

1. Average standardized values (this feels sketchy)
2. Limit the averaging to just the four "GDP-like" variables. Even then it's a little odd because it's averaging like GDP per capita PPP with nominal GDP per capita with GNP per capita, etc.

[ijyliu]

Hmm... I noted this earlier: https://github.com/ijyliu/ECMA-31330-Project/issues/44#issuecomment-843419217

So it looks like rescaling and then doing the mean is the way to go.

[ijyliu]

Here are the regressions and the start of a write up.

I think it would be good if someone would review my code at some point, particularly the bit on the means because that regression is behaving oddly relative to the simulations...

[ijyliu]

When I switched to using just the GDP-like measures the mean now performs almost identically to the first PC

[ijyliu]

@marionoro this looks kind of weird, and I don't know if it's better than what we had before. it also takes up a ton more space.

once we decide on that we can close this issue

[paul-opheim]

I think we should keep it like this so that it matches the terminology from the simulation tables. Could you apply \scalebox{0.75} so that it matches the simulation tables?

Also I think it looks odd partially because there is more spacing (between each line) than there was before. Do you have any idea why that is and how we could change it?

[ijyliu]

will see about the scalebox

The spacing was intentional. If we don't do it it looks like this, which is worse

[ijyliu]

Can do this though:

[ijyliu]

If this is good

we can close this issue

[ijyliu]

I think the title seems a little big (same applies to your tables too). I think we want to put the scalebox around the caption block also?

Or maybe we don't really need scalebox for the regression tables. They aren't that big anyway.

[paul-opheim]

I like the additional line, so let's keep that change that you made.

I agree that the title is a big big; I was having trouble getting the .tex to work when I included the caption in the scalebox though? I think the current scale (of 0.75) is good, just so that we keep our page count down as much as possible.

[ijyliu]

should probably scale this

also

[ijyliu]

^ Above has been scaled down, closing this issue.