ANOVA  interaction 为关键词，在官方SPSS搜索的
http://www.ibm.com/Search/?q=ANOVA+++++interaction&co=us&lo=any&ibm-submit.x=10&ibm-submit.y=21&sn=&lang=en&cc=US&en=utf&hpp=

交互显著后，事后检验分析 注意：主效应显著后，进行了事后检验分析
摘自：https://www-304.ibm.com/support/docview.wss?uid=swg21476140  
Problem(Abstract)
I have run a repeated measures ANOVA in SPSS using GLM and the results reveal a significant interaction. How can I use GLM to interpret the meaning of the interaction?

 For example, suppose that a researcher is interested in studying the effect of a new medication. To do so, she compares the effects of both the medication and a placebo over time. In this simple model, the finding of a significant Time X Treatment interaction means that the effect of time depends on whether the subject received the new medication or the placebo. Conversely, the interaction also means that the effect of treatment depends on time.

let's assume that there were two levels of time, weeks 1 and 2, and two levels of treatment, placebo and new medication.

One set of simple effects we would probably want to test is the effect of treatment at each time. Let's say we found that the placebo and new medication groups were not significantly different at week 1, but the new medication group was doing significantly better at week 2. These simple effects tests would support the assertion that the groups were equivalent at the start of the experiment and the new medication resulted in the difference observed at time 2.

Consider the hypothetical example, discussed earlier. Let's call the within-subjects effect Time and let's use the eight-letter abbreviation Treatmnt as the name of the between-subjects effect. To run the analysis and get tests for the simple effects of Treatmnt at each level of Time insert the following command syntax into the set of commands generated from the GLM - Repeated Measures dialog box.

/EMMEANS = TABLES(Time*Treatmnt) COMPARE(Treatmnt) ADJ(LSD)

Note that the optional keyword ADJ allows the user to specify anadjustment to the p-values for each set of pairwise comparisons which accompany the tests of simple main effects. The default adjustment is LSD, but users may request Bonferroni (BONF) or Sidak (SIDAK) adjustments.

Note that the EMMEANS subcommand allows specification of simple effects for any type of factors, between or within subjects. Thus if both factors were within-subjects factors (or between-subjects factors) the structure of the EMMEANS subcommand specifications would not change.


This brief sample command syntax file reads in a small data set and performs a repeated measures ANOVA with Time and Treatmnt as the within- and between-subjects effects, respectively. Two sets of simple effects tests are produced. The first is the effect of Treatmnt within each level of Time and the second is the effect of Time within each Treatmnt.

*The command syntax begins below.

data list free
/ treatmnt week1 week2 .
begin data
0 1 2
0 1 1
0 2 2
0 2 3
1 1 3
1 1 3
1 2 4
1 2 5
end data .

GLM
week1 week2 BY treatmnt
/WSFACTOR = time 2 Polynomial
/MEASURE = response
/METHOD = SSTYPE(3)
/PLOT = PROFILE( time*treatmnt )
/PLOT = PROFILE( treatmnt*time)
/EMMEANS = TABLES(treatmnt*time) COMPARE(treatmnt) ADJ(LSD)
/EMMEANS = TABLES(treatmnt*time) COMPARE(time) ADJ(LSD)
/CRITERIA = ALPHA(.05)
/WSDESIGN = time
/DESIGN = treatmnt.

这里我认为：
/EMMEANS = TABLES(treatmnt*time) COMPARE(treatmnt) ADJ(LSD)得改为：
/EMMEANS = TABLES(time*treatmnt) COMPARE(treatmnt) ADJ(LSD)  这样的描述统计与简单效应的结果一致。
这个事后检验呢，比较的是：treatment 因素的各个水平在time的各个水平；或者time的各个水平在treatment各个水平上  是否有显著差异！

咨询发表过ERP文章的在科研单位工作的学者（心理学实验功底扎实），她告诉我：上述的这个语句即适用于被试内变量也适用于被试间变量。

======================================================================================

交互显著后，SPSS如何进行简单效应分析
简单效应呢，与上述的事后检验不同的地方是：它比较的是某因素在另一个因素的各个水平上是否有显著差异。
如下主要内容都是源自舒华老师的那边书籍 之后水滴论坛SISI滴友的细致分享：
http://home.52brain.com/thread-7068-1-1.html

-----------------------------------
当两个或多个因素之间出现交互作用时，就需要进行简单效应检验。

比如在两因素2A(a1,a2)*2B(b1,b2)的实验设计中，如果A和B之间有交互作用，可以用简单效应来查看A因素在B因素的各个水平上（b1,b2）的处理效应，或者B因素在A因素各个水平上（a1,a2）的处理效应。

SPSS里面需要自己写语句，但是挺简单的。举两因素完全被试内设计的例子，可以这样写（蓝色为语句）：

manova A1B1 A1B2 A2B1 A2B2  %列出因素结合的变量名
/wsfactor=A(2) B(2)  %表明完全被试内设计
/wsdesign=A within B(1)
          A within B(2).  %计算A因素分别在B因素两个水平上的处理效应，别忘了最后那个点-----------------------------------
这个算是简单简单效应检验
例如A和B是组内变量，C是组间变量，如果考察组内变量A在B和C某一水平结合上的处理效应，可以这样写：
比如可以把组间变量C想象成性别（男，女），想分别看对于男被试，A因素在B1上的处理效应是否有差异。
可以这样写（蓝色为语句）：
 manova A1B1 A1B2 A2B1 A2B2 BY C(1,2)
/wsfactors = A(2)B(2)
/wsdesin = A within B(1)
/design = mwithin C(1).
-------------------------------------------
这里也有网友汇总了：
http://www.jky.gxnu.edu.cn/teacher/sitecss/1/index.asp?page=2&sitename=weiyiping&lmucat=��ݴ���

另外，我以前也有汇总，如下：

一。两因素完全随机，如果出现交互效应，进行简单效应比较，具体如下：
（一）
MANOVA dancen BY wendu(1,3) time(1,3)
 

咨询发表过ERP文章的在科研单位工作的学者（心理学实验功底扎实），她告诉我：上述的这个语句在被试内变量和被试间变量的描述是不同的，在SPSS编写语句时一定要注意！

舒华的书籍提及：在一个两因素完全随机实验中，可做简单效应检验有两组：一组可检验A因素在B因素的每个水平上的简单效应；另一组可检验B因素在A因素的每个水平上的简单效应。一般不需要同时两组都做，但选哪一组做，则与实验的理论假设和对实验结果的解释有关，即据实验的具体情况而定。
此外，这样简单效应检验只能得出处理效应显著或不显著的结论，它意义是什么？差异方向如何？还要通过图解来了解。-----这里通过对事后检验的了解，愚以为，上述的事后检验可以对简单效应的检验其实与图解效果相似。

咨询发表过ERP文章的在科研单位工作的学者（心理学实验功底扎实），她告诉我：在ERP分析中，只要把被试内和被试间变量定义对了，采用哪种相应的分析方法都可以。（即交互效应后的事后分析，简单效应分析都可以）

======================================================================================

编程算A因素在B因素各个水平的平均值标准差
转自：http://psychweb.psy.umt.edu/denis/datadecision/front/stat_II_2011/psyx_521_repeated_measures_bw.pdf
蓝色为编程内容 即：A因素在B的四个水平下的平均值与标准差：
/CELLS MEAN COUNT STDDEV.
http://s12/bmiddle/001pAHyRzy6HsI27Fx18b&690

http://s4/bmiddle/001pAHyRzy6HsI4sZTZe3&690

http://s10/bmiddle/001pAHyRzy6HsI6ljTz79&690

============================================================================================

具体的可以参见：
http://pan.baidu.com/share/link?shareid=145001&uk=3274657486《心理学研究方法》舒华张亚旭  进行了详细介绍。

选择LSD;和(BONFERRONI)的结果分析sig值有差异；


这个待有空再细致查阅总结，先写自己的经验总结：
20131222今天进行统计分析：重复测量，一个因素8个水平。事后分析：选择LSD;和(BONFERRONI)的结果分析sig值有差异，后者更加严格。当我两两进行配对样本T检验时，发现：LSD结果与配对样本T检验的一样的。
他们的差异多大呢，我举个例子，被试31人进行的分析
        LSD  .619   .002 
 配对T检验   .619   .002
Bonferroni   1     .044

官方的解释：http://www-01.ibm.com/support/docview.wss?uid=swg21475404

The three adjustments available for the significances which are available in EMMEANS are:

LSD(none), which is Fisher's Least Significant Differences, or unadjusted probabilities;

Bonferroni, which simply multiplies the significance by the number of comparisons (with a maximum significance of 1). Bonferroni will be too conservative if more than a few comparisons are made;

Sidak. Since there will often be too many comparisons for Bonferroni to be effective, this is often the only viable option if any adjustment to the probabilities is desired.

There are other adjustments, for example Tukey or Scheffe, which are valid for two-way interactions in a univariate analysis of variance. Currently, EMMEANS supports only the above three because they are also valid for models which include covariates, and in repeated measures models, both for main effects as well as for interactions which might mix between-subject and within-subject factors

================================================================================================