search

ajdamico / convey

variance of distribution measures estimation of survey data
GNU General Public License v3.0
17 stars 7 forks source link

potential edits to svyqsr text section? #373

Closed ajdamico closed 1 year ago

ajdamico commented 1 year ago

  1. the vardpoor library only allows one alpha parameter (https://rdrr.io/cran/vardpoor/src/R/linqsr.R), does that mean it's not possible to calculate the palma ratio with it?

  2. if i'm reading things correctly, i think the palma ratio would use alpha1 = 0.1 and alpha2 = 0.6 rather than alpha2 = 0.4. is that correct? i think using 0.4 instead of 0.6 would be an easy mistake for users to make, it might make sense to list those parameters directly in the text?

guilhermejacob commented 1 year ago

Ok, let's see:

  1. No, you won't be able to compute the Palma ratio with those.
  2. There are actually two ways to compute the Palma ratio. One is using svyqsr and the other is using svylorenz (they won't match exactly because of different estimators, but are asymptotically equal). The Palma ratio is the top 10% divided by the bottom 40%. In svyqsr, this means alpha1 = .40 and alpha2 = .90. Remember that, inside svyqsr, alpha2 works with svyisq( upper = TRUE ); that means that we will compute the total and subtract the bottom alpha2% to get the top (1 - alpha2)%.

By changing just alpha1, what we have is "symmetric complements" (kind of). That is, top (1-alpha1)% divided by bottom alpha1%.

ajdamico commented 1 year ago

would you be willing to incorporate my questions and your responses into the text? i could try if you don't have time