Recursive Algorithm for L1 Norm Estimation in Linear Models
β Scribed by Khodabandeh, A.; Amiri-Simkooei, A. R.
- Book ID
- 118175921
- Publisher
- American Society of Civil Engineers
- Year
- 2011
- Tongue
- English
- Weight
- 168 KB
- Volume
- 137
- Category
- Article
- ISSN
- 0733-9453
No coin nor oath required. For personal study only.
π SIMILAR VOLUMES
Necessary and su cient conditions are given for the consistency of the L1-estimator ΓΏ(n) of the regression parameter ΓΏ in linear models with independent but possibly nonidentically distributed errors. The heteroscedastic case is treated as a particular case. The asymptotic normality of ΓΏ(n) is also
Consider the model yi = x~llo + e,, i = 1 ..... n. Under very weak conditions on the error distributions, it is shown that inf.u ]Z~nl~'x,I = ~c is a necessary condition for the weak consistency of a minimum Lt-norm estimate of ,8o, which cannot be further improved.