Plot predictions after xthreg

[Andreas Orland]

Hello Statalist,

I have experimental panel data und used xthreg to estimate a threshold function (with one threshold).

Here is the output (it looks good and somehow as expected):

I now would like to plot this piecewise function with 95% confidence intervals.

The command margins, at(wealth2=(1(1)209)) ignores the "kink" at 115.9 and only gives me a linear prediction.
(I put it then into coefplot, that looks good besides being linear...)

Do you know how I can create predictions that take care of the threshold after xthreg? Thanks!

[Scott Merryman]

Note, xthreg is written Q. Wang in Stata Journal 15.1:

SJ-15-1 st0373  . . . . . . . . Fixed-effect panel threshold model using Stata
        (help xthreg if installed)  . . . . . . . . . . . . . . . . .  Q. Wang
        Q1/15   SJ 15(1):121--134
        fits the fixed-effect panel threshold model

Since it appear that -xthreg- is calling -xtreg- after determining the threshold and generating a categorical variable (_cat) I believe you will need to construct this graph in parts:

webuse grunfeld,clear
xthreg invest, rx(kstock) qx(kstock)
//Threshold at 1099
margins ,at(_cat == 0 kstock==(500(100)1100)) saving(file1,replace)
margins ,at(_cat == 1 kstock==(1100(100)1500)) saving(file2,replace)
use file1,clear
append using file2
line _margin _at2

[Andreas Orland]

Thank you very much!