Bounds output from invertci is strange

[a-torgovitsky]

library("lpinfer")
source("~/R/x86_64-pc-linux-gnu-library/3.6/lpinfer/example/dgp_mixedlogit.R")

dgp <- mixedlogit_dgp()
set.seed(1)
data <- mixedlogit_draw(dgp, n = 2000)

lpm <- lpmodel(A.obs = mixedlogit_Aobs(dgp),
               beta.obs = function(d) mixedlogit_betaobs(d, dgp),
               A.shp = rep(1, nrow(dgp$vdist)),
               beta.shp = 1,
               A.tgt = mixedlogit_Atgt_dfelast(dgp, w2eval = 1, eeval = -1))

r <- invertci(f = fsst, farg = list(lpm = lpm, data = data))
print(r$lb)
print(str(r$lb))

[[1]]
[[1]][[1]]
[1] 0.1042175

List of 1
 $ :List of 1
  ..$ : num 0.104
NULL

I see no reason that r$lb should be such a complicated object, nor why it should be different than r$ub. Why not just return r$ci as a simple vector with two elements?

[conroylau]

Done! In fact, I have added a data frame in an earlier commit when I added the feature for invertci to consider the case with multiple tuning parameters. However, it was in a different format and in another name.

I have revised the data frame for the confidence intervals as follows:

• The first column is the significance level.
• The second to the (n + 1)-th columns refer to the n tuning parameters for the procedure.
• The last two columns refer to the lower and upper bounds.

The lb and ub objects are redundant now so they are dropped.

Thanks!

[a-torgovitsky]

Looks good now, thanks