invalid xmin value error for every regression

[rawls238]

No matter what regression I run, I get this error:

Error in optimize(weighted.checkfun, lower = min(cvec), upper = max(cvec),  : 
  invalid 'xmin' value

I can run the same regressions with quantreg and lm and this looks like a code error, not an issue with my data? Let me know if you have any thoughts

[bcallaway11]

Perhaps, I haven't checked this in a while, but would be happy to look through it.

Would you mind providing some more details like what function you are using. Also, if you could send me some code where these errors happen, it would help too. You can email me at brantly.callaway@uga.edu if that works better than posting here.

Brant

[rawls238]

this is the code I was running:

jt.rand <- ci.qte(avg_time ~ restricted_application, data=phone_treatment, probs=c(0.5), se=T, iters=10)

restricted_application is a factor variable that takes on 3 values.

[fkohrt]

Same for me, here is a fully reproducible example:

set.seed(42)
a <- rnorm(25, 1, 1)
b <- rnorm(25, 0, 1)


cbind(
  measurement = c(a, b),
  group = rep.int(1:2, c(length(a), length(b)))
) |>
  as.data.frame() |>
  qte::ci.qte(formla = measurement ~ group, data = _)

[bcallaway11]

I think I got this one figured out. In the example that @fkohrt sent, the treatment variable is group, but it is coded as 1 or 2. If you just change this to 0 or 1, then it seems to work. Here is an example:

# Example 1: same as in issue
# only changes: (1) slightly reformat code
# (2) code the treatment (here `group`) as being 0 or 1 instead of 1 or 2
set.seed(42)
a <- rnorm(25, 1, 1)
b <- rnorm(25, 0, 1)


df <- cbind.data.frame(
  measurement = c(a, b),
  group = rep.int(0:1, c(length(a), length(b)))
)

res <- qte::ci.qte(formla = measurement ~ group, data = df)
qte::ggqte(res)

# Example 2: make sure this works with covariates
set.seed(42)
a <- rnorm(25, 1, 1)
b <- rnorm(25, 0, 1)

df <- cbind.data.frame(
  measurement = c(a, b),
  X = rnorm(50, 0, 1),
  group = rep.int(0:1, c(length(a), length(b)))
)

res <- qte::ci.qte(formla = measurement ~ group, xformla=~X, data = df)
qte::ggqte(res)

Please let me know if that works for you. And, assuming this works, I'll make a not to try to make the error messages more informative.

Brant

[fkohrt]

Hi Brant, thank you for this fast answer – and this great package in general! This solves the problem for me, though of course I don't know whether this also was @rawls238's problem.