Abstract
This paper proposes an accurate confidence interval for the trend parameter in a linear regression model with long memory errors. The interval is based upon an equivalent sum of squares method and is shown to perform comparably to a weighted least squares interval. The advantages of the proposed interval lies in its relative ease of computation and should be attractive to practitioners.
Original language | American English |
---|---|
Pages (from-to) | 1894-1902 |
Number of pages | 9 |
Journal | Statistics and Probability Letters |
Volume | 78 |
Issue number | 13 |
DOIs | |
State | Published - 15 Sep 2008 |
Keywords
- asymptotic normality
- linear regression
- long memory
- ordinary least squares
- weighted least squares
EGS Disciplines
- Mathematics