Dear all,

I would like to exclude some influential observations in a Tobit regression. When using .regress I have used predict's dfits, but it does not work for .tobit. I found a post in the old Statalist discussing how to obtain dfbetas using -jackknife- (https://www.stata.com/statalist/arch.../msg00865.html). I have not been able to find a similar post for dfits. I would be very grateful for any help on either locating a post on dfits, or the code itself.

I have one additional question. I have been able to run the code for dfbeta from the post above, but I enter a problem when using stored results in the code. This is the code and result that I get,

. // compute dfbetas
. local xvars dv6q1

. qui jackknife _b, keep nodots: qui tobit wtp1 `xvars', ll(0)
. qui tobit wtp1 `xvars', ll(0)

. . display e(depvar)
wtp1
.
. foreach z of local xvars {
2. gen dfb_`z' = ///
3. (1/_se[`z'])* (e(depvar)_b_`z'-_b[`z'])/(e(N)-1)
4. }
unknown function *()
r(133);

If I replace "e(depvar)" with "wtp1" the code works. My question is why does - e(depvar) - not work when as - display - shows it is correctly saved and - e(N) - works within the code?

I am using Stata/SE 15.1 for Windows.

Thank you in advance

Henrik