1974sword
《公共管理中的应用统计学》人大出版,原书45页准试验设计小节里讲到,"准试验设计可以用确立因果关系的四个标准中的三个标准来评价(内部效度):共变、时间顺序、非伪关系。另一个不能用的标准是理论。"
我想不明白,为什么这里不能用理论来检验因果关系呢?准试验设计可以不符合现有理论吗?
请诸位高手予以解答???
联系邮箱:1974sword@163.com
yihui
我只想告诉你我对你重复发帖的行为感到愤怒。
排除这种情绪,我再告诉你,“因果关系”是哲学家的问题,若真要争辩,那么永远都说不清楚。上面总结的四个标准基本都可以视为废话,尤其是“非伪关系”。
1974sword
对不起,重复发贴是因为我第一次发错了地方,发到了SPSS应用中去,我认为这个问题不属于那一类,所以重发了一次,希望给点建设性意见,谢谢
yihui
还好你只用了三个问号。
发错了就不要重新发,自然会有人帮你转移。重复发比发错地方更低效。
1974sword
好的,知道了,谢谢你了!
不知有没有更好的解答?
anita_jiu
"因果关系”是哲学家的问题 - broadly agree with Xie.
"为什么这里不能用理论来检验因果关系呢" - I think the answer to this question is 'it depends'. Important elements may include philosophy position (that Xie has addressed), research strategy and design. Perhaps the philosophy position would be the primary element here in determining whether or not an existing theory can test causal relationships associated in an inquiry. This is, in general, a too broad question to give you a definite answer in sentences. It obviously covers a wide range of questions for consideration, I believe.
Internal validity - can be influenced by many elements, research design, sampling (probability or non-probability), measurement of variables (may be linked with theory and/or exploratory findings), measurement scales ... Theory may be considered as a general guide to understand a situation; however, there is no guaranteen that theory holds the absoute 'truth'. Otherwise why do [u]some [/u] researchers keep testing and/or refining theory? The main reason that I underlined 'some' is that there are some researchers who do not agree with this position.
Perhaps more reading on other research methods would be helpful? Good luck.
1974sword
谢谢anita_jiu及谢的答复,但问题是,书中认为,实验设计可以用这四个标准来确立因果关系,而准实验设计只能用前三个,而不能用理论,我不明白,为什么实验设计可以用理论来检验,而准实验就不能 呢?请各位高手再帮助我分析一下,有原书的话最好能看一下,更能理解我的问题,谢谢了。
anita_jiu
准实验设计 - this may be a primary question of your philosophy stance which drives your research by a number of core assumptions. Under the 'guideline' of these assumptions, a particular school of thoughts (e.g. positivism, critical realism, constructivism) has different 'favour' in terms of methodology strategies, research methods, data collection and so on. Experiment designs are useful for causal research which is particularly designed to test theories. Researchers can manipulate and control independent variables, record and investigate the changes of dependent variables. In so doing, researchers aim to find causal relationships between variables. However, experiment desings have different types to meet different research objectives. You could try the following references for a broad understanding of methodology issues in the marketing field.
1. Churchill G.A.Jr (1988), Basic Marketing Research, The Dryden Press, USA
2. Malhotra N. K. (2002), Basic Marketing Research: Applications to Contemporary Issues, Prentice Hall, USA
Apart from the above, I would also like to recommend you to explore more in the topic of philosophy. Without a sound understanding of various theoretical positions, an inquiry could suffer serious impacts later on. By the way, the discussion above is not targeted to answer your particular question as I thought your question somewhat too broad to answer. However, I suppose the question is inherently linked with many other issues such as philosophical stance, methodology choices. Therefore, I 'tackle' the iceburg (your question) from another perspective. Not sure whether this would help or not. Good luck.
1974sword
谢谢楼上的建议
yihui
举个简单的例子说明为什么统计学不能证明因果关系:拿最普通的回归来说,y = b0 + b1*x + error这样的数学式子完全可以移项得到 x = -b0/b1 + (1/b1)*y - error/b1,仅凭统计学和数学,鬼才知道谁是原因谁是结果。所以说统计学的优势在于定量化描述事物之间的关系,而因果的确定问题则在其界外,如Anita所说,那是Philosophy的问题。
1974sword
人大版的那本教材难道写错了?书是李静萍老师校译的, 我给她发过两封邮件求教,也没见回音,不知道什么原因,难道真象楼上讲的,书本来写得有问题?
anita_jiu
有问题的不见得肯定是书中的观点不正确。相反,我们首先要知道书中的上下文,方能客观地评价书中的观点。我没有看过这本书,不想断章取义给你所谓的`答案`。这是初衷。
再者,在很多调查里,错误本身就是无法避免的,但我们可以尝试将其降到最低或可接受程度。我认为,没有调查是完美的。做调查,有很多方方面面的选择要研究者去解决。这些选择少不免会有subjective and arbitrary。当然,我是主张to make subjective decisions based on rationales.
你问的问题,其实是有点`一言难尽`。因为这里面包含了好多因素和其他重要概念。我想,调查人员的philosophy position是其中最根本的原因。打个比方,如果我是a constructivist,那么你提的问题对于我来说根本就不是问题,因为我根本就反对to gain causal relationships through theory testing. Rather, a theory comes from the society which is structured by human beings while human beings construct theories in contexts individually. 当然,这不过是个例子,而且你可能会觉得我扯远了。我想说的是,哲学论完全影响我们调查事物的看法、调查方法、收集手法等。只是一般来说,它都被忽略了。但事实上它是很重要的一个概念。
统计和数学只是工具帮我们理解事情,但觉不是唯一的工具。我赞同(大、小)谢(?!又不算是学友,哈哈)学者的观点。如果你有兴趣的是公共关系管理学,那么它不是自然学科而是social science。这里面就包含了许许多多无法控制的东西(甚至我们无法解释的东西),如人及其行为。在管理学里,理论也不一定是完全正确的,我们也见过有理论可以或不可以解释某种事物。但这不等于说理论肯定是错误的。要客观,全面分析才应该perhaps tentative suggestion rather than conclusion.
1974sword
谢谢anita_jiu的解释。
我觉得有必要再重申一下问题:
书的作者认为:
在实验设计中, 为了确保试验效度,可以用四个标准来确立因果关系,即共变、时间顺序、非伪关系及理论,而在准实验设计一节只中,作者认为检验内部效度的标准只能用前三个,而不能用理论。
问题是,在作者那里,为什么实验设计可以用理论来检验,而准实验就不能呢?它们之间的有关效度检验中究竟存在哪些差异而使得理论标准在这里能用而在那里就不行了呢?唉,也不知李静萍老师对此如何理解,她应该非常熟悉所指的那本书。
yihui
她的sohu邮箱有时候收不到邮件,你试试Gmail,用户名不变。
netcow
我不同意因果关系是哲学家的问题,相反真正能确立或解决因果 关系应该是各个具体科学领域的科学家们。英国关系的确立应当而且必须引入概率的观点。至于统计学关系与因果关系,应该是首先由统计学关系,然后才可以谈因果关系。流行病学里面,确立一种危险因素(现代意义上的广义的病因,凡是能够使得疾病发生概率增加的因子都可以称之为危险因子,或看成是该种疾病的病因)与某种疾病的因果关系有9条标准,但并非9条标准均满足,才是病因,但必须要满足的是联系的顺序,即因在前,果在后 。其次是联系的强度,在是联系的可重复性。原因与结果的指征之间这些联系的原则成立的越多,判定为因果关系正确的把握就越大。
1974sword
确立因果关系的确不是一件易事,但这方面的工作我们还是可以做一些,我认为并不是在因果关系上我们就是完全的无能为力的,只能交由哲学的思辩以及无何止的争论。
希望再有同行探讨本贴中提出的问题,谢谢了
anita_jiu
[quote]引用第14楼netcow于2007-03-30 10:59发表的“”:
我不同意因果关系是哲学家的问题 - I think I emphasised 'the philosophy position' which leads to your point - '相反真正能确立或解决因果 关系应该是各个具体科学领域的科学家们/researchers who are fundamentally influenced by their own position'。
1974sword
谢谢大家的讨论