you can. just w/ lots of side effect like natural heteroskedasticity

送交者: mangolasi 于 2007-04-18, 16:16:26:

回答: 一般回归分析方法 can not be used because response (+/-) is binary 由 HunHunSheng 于 2007-04-18, 15:56:35:

or w/ predicted value outside [0,1] etc...regression is not only for prediction but also for seeing source of variation or causality etc.

nonparametric stuff like fisher's exact can be useful in getting source of variation for discrete explanatory variables, but they save you from misspecification.

所有跟贴:

Traditional regression does not make any sense in this case (无内容) - HunHunSheng (0 bytes) 2007-04-18, 16:22:42 (136377)
- where? I saw people who have to use traditional regression for binary - mangolasi (241 bytes) 2007-04-18, 16:26:18 (136382)
  - No friggin expected value for binary response - HunHunSheng (32 bytes) 2007-04-18, 16:28:59 (136383)
    - you can do a simple linear regression with binary response - xj (43 bytes) 2007-04-18, 18:05:39 (136393)
we should call for the lost in xlogit to explain this... (无内容) - mangolasi (0 bytes) 2007-04-18, 16:18:41 (136375)

加跟贴