关于ReHo，ALFF的统计检验问题

Submitted by lwz110911 on Mon, 10/29/2012 - 19:45

大家好：
在用dparsf软件得到ReHo，ALFF后，我用rest做了双样本t检验（对照组的smReHo，抑郁组的smReHo。对照组smALFF，抑郁组的smALFF）我有几个问题想问问大家。
               1   这样做出来的结果具有统计学上的显著差异么？   如果没有请问为什么呢？   
               2   假如我把对照组的每一个人和抑郁组每一个人进行双样本t检验，这样做出来的结果好么？如果不好为什么？
               3   如果我说的这两种方法都不可行，希望大家能给我些建议。
               谢谢大家

Submitted by YAN Chao-Gan on Mon, 10/29/2012 - 22:58

1. 做对照组与抑郁组关于smReHo的双样本T检验是可行的。如果没有显著差异，可能性有很多：样本量，样本，年龄性别等其他控制因素。
2. 这种方法不可行。双样本T检验不能做一个被试VS一个被试。

Submitted by lwz110911 on Tue, 10/30/2012 - 11:03

首先谢谢您的答复，我自己也做了下抑郁组、对照组的双眼本t检验 结果如下（不知道这2个图片是否有用）
                                  1   这幅图像是正常人与病人ReHo差异的图片。
                                  2     这幅图像是正常人与病人alff差异的图片
                                请问您，从这两幅图像中能否看出二者的显著差异。我觉得是可以的。
                                 谢谢！（我是个学生 有可能提出来的问题有点弱，请您见谅。）

Submitted by Qing Zhao on Thu, 11/01/2012 - 01:58

First, be clear about what it means when you say "significant"

There are only 2 samples from 2 different huge sets, and so you never know if the mean of these two sets are different or not. Statistical index can be calculated from these 2 samples and help you to decide the possibility of it (the 2 sets have different means or not). And if the possibility is high, we call it "significant difference".
 
So statistic never tell you that "the healthy controls are significant different from patients", it just about numbers not people. But for each voxel, there is a ReHo or ALFF value, and statistic can tell you if these values from healthy controls seems different from those from patients or not.