SETAR confidence intervals and standard errors in tsDyn

[Lukas Wiechers]

Dear Matthieu Stigler,

I have estimated a SETAR model using your tsDyn package and then done a
forecast for ten time periods using the predict.nlar command. The code
is running, however, I am not sure about the received standard errors.
Given the estimated SETAR model with 229 observations named
"mod.setar21" I run the following code:

predict(mod.setar21, n.ahead=10, type="MC", nboot=100, ci=0.95)

According to the manual (p. 45), "$se" are the estimated standard
errors, but I get negative values as seen in the following:

$se
Time Series:
Start = 230
End = 239
Frequency = 1
5% 95%
230 -0.04150263 -0.04150263
231 -0.10061752 0.02058175
232 -0.11711327 0.05139648
233 -0.12885112 0.06643442
234 -0.13405834 0.08175834
235 -0.14099120 0.08076885
236 -0.15169921 0.10232657
237 -0.14558111 0.10146933
238 -0.16921491 0.10573619
239 -0.14088897 0.10201412

Since these values change if I change the value of "ci" in the predict
command and the two quantiles then adjust as well to the selected
confidence level, I want to ask you if these values are already the
lower and upper bound of the estimated confidence interval.

Thanks in advance!

Best regards
Lukas Wiechers

[Matthieu S]

Hi Lukas

Well spotted! Indeed the output is a confidence interval, not the se. I will update this later on, either changing the doc or the function. So stay updated. I've opened an github issue for that.

Also, note that there is a new published paper describing tsDyn link, I would appreciate if you can refer to this one for citation purpose.

Best,

Matthieu

--
You received this message because you are subscribed to the Google Groups "tsdyn" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tsdyn+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tsdyn/614aaa85-08f8-fbf8-70e5-660e0fc96e37%40mail.upb.de.