h*i 发帖数: 3446 | 1 推荐一本书,是对概率论的"新解":E. T. Jaynes ,Probability Theory: The
Logic of Science, Cambridge University Press,2003
Free: http://omega.albany.edu:8008/JaynesBookPdf.html
Amazon: http://www.amazon.com/exec/obidos/tg/detail/-/0521592712/
我们学的概率统计[1],是这本书的作者Jaynes所攻击的所谓"正统"的概率统计。如果是
学生物,社会科学出身的,肯定学的是R.A. Fisher的理论。记得我们当时学心理统计
学的时候,用的书是系里一个退休老先生,据说是Fisher的弟子写的,油印本,两大册
,文白夹杂,很古雅,基本上是系里的祖传密籍的地位(当然,我们更热衷于八卦老先
生的法国老婆,呵呵)。这本书把Fisher描述成一个学霸,对他的理论和为人大大地抨
击了一番,认为他的理论延迟了科学,发展呵呵。如果学数学物理出身的,可能学概率学
得比较深,就会从测度论来讲。这本书对测度论的无 | A***A 发帖数: 98 | 2 I doubt. It'll be a while before books like this one (and IMHO there are
much better ones, more balanced, with more content and less bitterness) make
any kind of a dent on the way statistics is taught in psychology programs
across this country.
I am not a frequentist, and I use Bayesian computational methods on a daily
basis. The probability theory that I am most familiar with is the
Kolmogorov system, which, IMHO, is well-grounded on the real-line measure
theory of Lebesgue and should simply | h*i 发帖数: 3446 | 3 老实说,我对贝叶斯数据分析的方法的普及也不抱任何幻想。美国一般学心理学的数学
能力太差了,不可能一下子就变了。只可能慢慢的变。不过我觉得在中国,要有好老师
,是可以迅速打开局面的。
但我说的革命性的影响,更多的是从对认知科学的影响。 作认知心理学的可能不少都
有一种体会,就是传统心理学实验方法感觉走到绝路了,要想往神经生理上靠,但怎么
比得过人家学生物的呢?没有自己的一套东西是不行的。我觉得,贝叶斯的观点可能是
拯救认知心理学的一个出路。事实上,这个革命已经开始了,老家伙们支持,新AP们都
在往贝叶斯的路上靠。 看看现在Bulletin, Review的文章就知道了。
Jaynes的观点对心理学的价值大家还没有完全认识到,我觉得是很有可挖掘的。他的
bitterness其实也正常啦,在学术界是算一般的。冷嘲热讽的文章比这厉害的多多了。
大概你们搞psychometrics的比较civil吧,呵呵。
至于Kolmogorov的概率系统,Jaynes自己也认为和贝叶斯的是一致的。Kolmogorov的系
统是好的数学,没有人否定这个,问题是和现实没有关系。好的数学系统多了去了,能
描述现实
【在 A***A 的大作中提到】 : I doubt. It'll be a while before books like this one (and IMHO there are : much better ones, more balanced, with more content and less bitterness) make : any kind of a dent on the way statistics is taught in psychology programs : across this country. : I am not a frequentist, and I use Bayesian computational methods on a daily : basis. The probability theory that I am most familiar with is the : Kolmogorov system, which, IMHO, is well-grounded on the real-line measure : theory of Lebesgue and should simply
| h*i 发帖数: 3446 | 4 Care to recommend some? I am very interested.
I have Gelman's "Bayesian Data Analysis". I like it. Any others that you can
recommend?
..................
【在 A***A 的大作中提到】 : I doubt. It'll be a while before books like this one (and IMHO there are : much better ones, more balanced, with more content and less bitterness) make : any kind of a dent on the way statistics is taught in psychology programs : across this country. : I am not a frequentist, and I use Bayesian computational methods on a daily : basis. The probability theory that I am most familiar with is the : Kolmogorov system, which, IMHO, is well-grounded on the real-line measure : theory of Lebesgue and should simply
| r******s 发帖数: 2155 | 5
make
Examples?
【在 A***A 的大作中提到】 : I doubt. It'll be a while before books like this one (and IMHO there are : much better ones, more balanced, with more content and less bitterness) make : any kind of a dent on the way statistics is taught in psychology programs : across this country. : I am not a frequentist, and I use Bayesian computational methods on a daily : basis. The probability theory that I am most familiar with is the : Kolmogorov system, which, IMHO, is well-grounded on the real-line measure : theory of Lebesgue and should simply
|
|