ͽÄêɽ -¾ÜºÙ¾ðÊó-
·ï̾ | LS ÅÄÃæµ×Ì»á |
³«»ÏÆü»þ | 2008ǯ 10·î 28Æü (²ÐÍËÆü) 12»þ00ʬ (GMT+09:00) |
½ªÎ»Æü»þ | 2008ǯ 10·î 28Æü (²ÐÍËÆü) 12»þ45ʬ (GMT+09:00) |
¾ì½ê | 26¹æ´Û 7F¡¡702¶µ¼¼ |
¾ÜºÙ | Êó¹ð¼Ô¡§ÅÄÃæ µ×Ì (Áá°ðÅÄÂç³Ø À¯¼£·ÐºÑ³Ø½Ñ±¡) ȯɽ¥¿¥¤¥È¥ë¡§The Isotonic-Regression Based Least Squares Estimation of Semiparametric Linear Index Models ³µÍס§ Isotonic Regression (IR) is a nonparametric method to estimate functions by direct minimization of mean squared errors over a functional set with shape restrictions. The IR is fast¡¤ efficient and free from tuning parameters such as kernel bandwidths. In this paper¡¤ a new class of the IR-based estimators for the single index model is proposed. The class includes the Semiparametric Least Squares (SLS) estimation¡¤ the Semiparametric Method-of-Moments (SMM) estimation¡¤ and the Iterative Least Squares (ILS) estimation. The paper mainly investigate large sample properties of these estimators¡¤ in particular¡¤ the cubic convergence rate of the SLS estimator¡¤ root-n asymptotic normality of the SMM estimator¡¤ and near semiparametric efficiency of the ILS estimator. JEL classification: C14; C25 KEYWORDS: Single Index Models; Binary Choice Models; Isotonic Regression; Semiparametric Least Squares Estimation; Iterative Least Squares Estimation ¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦¡¦ ¤³¤Î¸¦µæ²ñ¤Ï»²²Ã»ñ³Ê¤ò¸ÂÄꤷ¤Ê¤¤¥ª¡¼¥×¥ó¤Ê¤â¤Î¤Ç¤¢¤ê¡¢³§ÍͤÎÀѶËŪ¤Ê¤´»²²Ã¤ò¤ªÂÔ¤Á¤·¤Æ¤ª¤ê¤Þ¤¹¡£ ÅöÆü¤ÏÃë¿©¡Ê¤´ÈÓÊÛÅö¡Ë¤ò¤´ÍѰդµ¤»¤Æ¤¤¤¿¤À¤¤Þ¤¹¡Ê¸ÂÄê40¿©¡Ë¡£ ¤ª°û¤ßʪ¤Ï³Æ¼«¤´»ý»²¤¯¤À¤µ¤¤¡£ |
¥«¥Æ¥´¥ê¡¼ | À¯¼£·ÐºÑ³Ø´ðÁÃ¥»¥ß¥Ê¡¼¡¦¥é¥ó¥Á¥¿¥¤¥à¥»¥ß¥Ê¡¼ |