How to create toy data with autocorrelation?

[Mike Smet]

Hi,

I want to create a fictional dataset where the error term of a time series regression should be characterized by p-th order autocorrelation.
I'm stuck with imposing the autocorrelation ... Does somebody has a clue how to generate this type of data?

Below you find something similar where I generated heteroskedasticity (which turns out fine) ... so I would be very happy to be able to create something similar, but with autocorrelation ...

clear
set seed 123
set obs 200
* Creating independent variable
gen x=runiform(1,1000)
* Creating heteroskedastic resid
gen sd=runiform(0,4)
gen u=2*rnormal(0,sd*x)
* Creating dependent variable
gen y=1000+10*x+u
* Regression (and plots indication heteroskedasticity)
reg y x
scatter y x
rvfplot

Thank you very much!