This preview shows page 1. Sign up to view the full content.
Unformatted text preview: s not achieved. t5 distribution replicates
heavy tails of residuals quite well (based on the histogram),
therefore we may use quantiles of this distribution for obtaining
prediction intervals:
Xt +h ± tα/2,5 · s.e.(Rt ),
where Xt +h – point forecast; Rt – estimated residuals.
In R, tα/2,5 can be obtained by:
> alpha < 0.05
> qt(alpha/2, df=5)
[1] 2.570582
Compare with normal distribution:
> qnorm(alpha/2, 0, 1)
[1] 1.959964
GARCH model for DAX time series Other approaches: QQ plot Instead of using the histogram superimposing, the model distribution
for residuals may be selected using QQ plot.
However, in R some extra coding is required to compare the
quantiles from observed data with those from hypothesized
distribution if the latter is not normal. The function qqnorm is only
for normal QQ plots. GARCH model for DAX time series Other approaches: bootstrap
Alternatively, we may bootstrap the residuals (sample with
replacement many times) and obtain the quantiles f...
View
Full
Document
This note was uploaded on 08/04/2013 for the course ECON 201 taught by Professor Vandewaal during the Spring '09 term at Waterloo.
 Spring '09
 VANDEWAAL

Click to edit the document details