search

gphk-metrics / stata-multe

Multiple Treatment Effects
17 stars 4 forks source link

Pull Request for #1: Translate #2

Closed jerrayc closed 2 years ago

jerrayc commented 2 years ago

Complete translation (#1 ) + additional option to save tauhat and lambda

mcaceresb commented 2 years ago

@jerrayc Should the names for lambda/tau start at 1? I thought the coefficients corresponded to estimates from the second level onward.

jerrayc commented 2 years ago

@jerrayc Should the names for lambda/tau start at 1? I thought the coefficients corresponded to estimates from the second level onward.

Ah I see what you mean. I think it's better to start indexing with 1, since there does not exist a tauhat for level k = 1. If we want to change our numbering of k to more clearly parallel our indexing starting at 1, I would be down to make those adjustments.

In practice we don't use the level k=1 since it is the control group, so in my head I index that as 0, while the rest of the treatments start with 1, 2, etc. I think this is more in line with what the theory note writes and is less confusing, since there does not exist a tauhat for k = 1 and it seems weird to start indexing at 2. Does that make sense?

mcaceresb commented 2 years ago

@jerrayc Should the names for lambda/tau start at 1? I thought the coefficients corresponded to estimates from the second level onward.

Ah I see what you mean. I think it's better to start indexing with 1, since there does not exist a tauhat for level k = 1. If we want to change our numbering of k to more clearly parallel our indexing starting at 1, I would be down to make those adjustments.

In practice we don't use the level k=1 since it is the control group, so in my head I index that as 0, while the rest of the treatments start with 1, 2, etc. I think this is more in line with what the theory note writes and is less confusing, since there does not exist a tauhat for k = 1 and it seems weird to start indexing at 2. Does that make sense?

Ok. I defer to you here. Feel free to squash and merge if there's nothing more on your end. Thanks!

mcaceresb commented 2 years ago

Summary here.