h********F 发帖数: 484 | 1 怎么办?
本来student t-test是significant的,但用anova p就大于0.05了。。。 |
|
D*a 发帖数: 6830 | 2 两组数据你为啥想起来用anova?
ps P应该是一样的 |
|
k*****d 发帖数: 12 | 3 如果只有两组数,Anova是假设equal variance的t-test。大多数的软件的缺省的t-
test是unequal variance的t-test。 |
|
i****t 发帖数: 58 | 4 你是否用了one-tail t test? t test 的pvalue 是否是anova 的一半 |
|
|
|
r*****5 发帖数: 1876 | 7 谢谢大家,
Two way ANOVA and multiple comparison给出来的结果看不懂啊,感觉不是我想要的
那两个比较 |
|
r*****5 发帖数: 1876 | 8 谢谢你:)
分开做的话,是用ONE WAY ANOVA 就可以了,对么
可是分开的话,就是两两比较,我怎么觉得分开的话用T-TEST 就可以了啊?
, |
|
D*a 发帖数: 6830 | 9 one way ANOVA 然后post hoc也可以的吧,看见很多文章这么发啊,还是不错的杂志。 |
|
D*a 发帖数: 6830 | 10 我个人觉得你这个问题就是各组之间有没有区别而已,到底AB之间如何作用是作出的结
论,用one way ANOVA 我觉得没什么不对的。post hoc 是两两比较但是alpha值是修正
过的,因此不如student t 敏感,这样来避免统计中的一类还是二类错误。post hoc
也有很多常用的是bonferroni 。手机上网实在不方便,你google下我说的这几个关键
词吧。
但是如果临床治人的实验一般统计有专人把关,我说的不一定够。搞老鼠什么的我觉得
知道调alpha就差不多了,当然我说的不是那种牵扯很多bioinfo的课题。
我统计也是半吊子,等大牛吧 |
|
o********r 发帖数: 775 | 11 Two way ANOVA will give you 3 p values:
One for each of the main effects and one for the interaction term (whether A
effect is independent of B effect).
Well, after a second look, I am not sure what is your question. You may
have asked whether B has effects (regardless of A). If this is true, check
the p value for main effects of B.
Actually you just need to run linear model or GLM to estimate the p values
in your setting and you can specify the proper model (outcome~A+B or outcome
~A*B) base... 阅读全帖 |
|
发帖数: 1 | 12 you can model interaction with multiple linear regression in R
for example, y~a+b+c assumes no interaction, and
for interaction, use y~a+b+c+a:b+a:c+b:c+a:b:c
a:b/a:c/b:c are two-way interaction terms
a:b:c is the three-way interaction term
if a/b/c are continuous, t-test in summary() should be enough to see if the
interaction terms are significant. if categorical, use Anova(). |
|
B****n 发帖数: 11290 | 13 ANOVA=A NOVA=a cataclysmic nuclear explosion caused by the accretion of
hydrogen onto the surface of a white dwarf star |
|
C****o 发帖数: 994 | 14 use Excel. Just type in ANOVA in the "help" and then the instruction will
show up. |
|
p********a 发帖数: 5352 | 15 ☆─────────────────────────────────────☆
chuyun (chuyun) 于 (Tue Apr 3 23:40:24 2007) 提到:
我有5个factor,要做ANOVA,
老师要求首先fit full model,
包含5个main effects,10个2-way, 10个3-way,5个4-way和1个5-way的interaction.
结果发现5个4-way interaction中有4个的Type III SS 和 df 都是0。
(我用SAS和SPSS都试过,结果一样)
我的sample size大约是950,
5个factor分别有2,2,2,3,4个level,
所以df=0不可能因为样本量不够。
老师问我那到底是什么原因造成SS 和 df 都为0,
实在是想不出来,
不知道各位有什么意见没有,
谢谢!
☆─────────────────────────────────────☆
zyyz789 (wokao) 于 (Wed Apr 4 00:06:07 2007) 提到:
only consider |
|
m*****8 发帖数: 654 | 16 在研究anova 和ancova的理论。想找treatment effect的point estimate, variance
和testing的公式,但是找不到很气馁。有没有大侠可以推荐一下文章或者书可以看看
。 |
|
S****y 发帖数: 1342 | 17 在ANOVA post hoc里面
3组同源的数据,但是用不同的方法处理,处理后 我要测试他们 是不是差异显著,该
用什么方法?
谢谢,包子奉上 |
|
n*a 发帖数: 12 | 18 for a given ANOVA study involving three treatments, with n1=n2=n3=9, it was
found the MSE=18. the appropriate estimeated standard deviation of the
sampling distributionof x1bar-x2bare to construct a confidence interval for
u1-u2 is 4
判断题,正确请给出理由
错误也请给出理由 |
|
m**i 发帖数: 37 | 19 比如说有三个group,对于每个group,我知道平均值,uncertainty for 95%
confidence level, 和degree of freedom (不是整数)。现在比较平均值,还能用one-
way ANOVA 吗?怎么计算F值?
谢先。 |
|
m**i 发帖数: 37 | 20 "degree of freedom (不是整数)": My case is a little different, not like the
general case: we measure one quantity n times in group-1, m times in group-
2, etc. But I calculate the mean of the quantity for each method, the error
(uncertainty) including random and systematic errors, and the degree of
freedom which is in general not integral (not n-1. m-1. etc).
So in my situation, it is still one-way, different degrees of freedom for
each method. what is proc anova?
Thanks. |
|
D*a 发帖数: 6830 | 21 我有Tg老鼠和control老鼠两组,我把它们两两交配之后形成四组, 测后代的某个值,
我用SPSS上面的anova之后,再选什么test两两比较呢? 看着SPSS上面的一堆test眼晕
。。。
大侠指点一下吧,多谢了 |
|
|
b*****e 发帖数: 223 | 23 看这个: from
http://faculty.chass.ncsu.edu/garson/PA765/anova.htm
The reason for this assumption is that the denominator of the F-ratio is the
within-group mean square, which is the average of group variances taking
group sizes into account. When groups differ widely in variances, this
average is a poor summary measure. Violation of the homogeneity of variances
assumption will increase type I errors in the F test (wrongly rejecting the
null hypothesis). The more unequal the sample sizes in the cell |
|
b*****e 发帖数: 223 | 24 借帖子问个弱问题,我有点糊涂了,上学时候那本书教的是 linear mixed effects
model。LME 和 ANOVA 是什么关系? |
|
h******e 发帖数: 1791 | 25 这个不用anova,你主要想比较treatment和control的平均值有没有区别,用t-test就
行了。 |
|
f*********e 发帖数: 45 | 26 anova 中算出的p值, 和t 检验 pool variance 算出的 p 值, 是一样的么? |
|
t********y 发帖数: 469 | 27 比如说
sampling或者DOE 得到一组数据
有变量 y,x,class,s
y 和 x 都是continous variable
class和 s是 categorical variable
那么用proc surveyreg还是用anova?
来做这个model呢? |
|
|
l*********s 发帖数: 5409 | 29 aov merely extracts ANOVA table from the lm object. Purely for cosmetic
purpose. |
|
s********9 发帖数: 74 | 30 Whne use ANOVA to test five groups's mean difference, the model shows
significant difference. While when use multiple comparasion method to
checked each group pairs, there is no difference between any two groups.
What is the problem? |
|
g**r 发帖数: 425 | 31 No problem, this can happen, can even happen just within ANOVA without MC
adjust
A simple example: three treatment, A, B, C. Model:
Y=A B
Global testing: A=B=C, reject
testing: 1. A=C failed to reject; 2. B=C failed to reject
The global testing is a much more powerful test, but the detailed pairwise
testing, even without splitting alpha or using family-wise alpha, is less
powerful. |
|
F****y 发帖数: 11 | 32 要用Repeated measure ANOVA 但不知道是啥。。。? 和mixed effect model有啥区别
没~
谢谢啦! |
|
A*******s 发帖数: 3942 | 33 不大懂你想问啥
先想这个问题
如果不是因为interaction,
为啥不用two one-way anova tests |
|
A*******s 发帖数: 3942 | 34 我不大记得anova的东西了。他说的是type I SS么?那一般的顺序确实是先main
effects再interaction。 但是他言下之意好像是,测不出main effect就不用测
interaction了?这个我没怎么听说过。
请大牛出来解惑吧。 |
|
h******3 发帖数: 190 | 35 和marginal means有什么区别?
在大于1 way的ANOVA
谢谢。 |
|
R*******c 发帖数: 249 | 36 wiki上关于model identifiability的定义是:
Mathematically, different values of the parameter must generate different
probability distributions of the observable variables.
在anova里,假如模型是: Y_ij=m+b_i+\eps_ij, b_i~N(0,var_b),
\eps_ij~N(0,var_eps)
我想问的是如何从定义去证明var_b和var_eps是identifiable的?因为好像从Y的
marginal分布
是只能确定var_b+var_eps的...
或者更普遍的情况是,在random effects model或者mixed effects model里如何证明一个模
型的identifiability呢?
多谢帮忙啊~ |
|
l*********s 发帖数: 5409 | 37 you are talking about fixed model ANOVA,right? If so, the means are unknown
but fixed, the estimation error is sigma_e/sqrt(n), so it is identifiable. |
|
R*******c 发帖数: 249 | 38 No, I am talking about random effects anova, because b_i~Nor(0,var_b),
so b_i is random.
Do you have any ideas?
unknown
identifiable. |
|
R*******c 发帖数: 249 | 39 不好意思啊,可能是我理解有问题
wiki里的关于identifiability的定义是说根据observation的分布能唯一确定模型中所
有参量
的值才能叫identifiable。
在这个random-effects anova里面,如果给了观测值,y,那么这个y的分布是
normal(0,var_b+var_eps).所以从y的分布我们只能确定var_b和var_eps的和,而不能
唯一
确定这两个值分别是什么啊
我不理解的是,我这样从定义去推,然后推出说这个模型不是idenfiable的,这里面哪
一步出了问
题啊? |
|
|
r*****o 发帖数: 140 | 41 不是特别明白你的意思。
不过为什么不把条件作为一个factor?在一个ANOVA下是不是好办一点?
你要看的似乎是条件和factor的interaction. |
|
s********g 发帖数: 889 | 42 我有一个two-way anova的结果:
Df Sum Sq Mean Sq F-value p-value
Mix type 3 0.056 0.019 19.9 < 0.001
Aggregate type 2 0.218 0.109 115.6 < 0.001
Interaction 6 0.014 0.002 2.5 0.054
Residual 24 0.023 0.001
结论是两个因素都对实验结果有显著影响,但我可不可以继续说Aggregate type对实验
结果影响更significant?因为他的F-value大很多。
谢谢 |
|
s********g 发帖数: 889 | 43 p值是这样的
Df Sum Sq Mean Sq F value Pr(>F)
Agg 2 0.218400 0.109200 115.6235 4.775e-13 ***
Mix 3 0.056322 0.018774 19.8784 1.077e-06 ***
Agg:Mix 6 0.013911 0.002319 2.4549 0.05402 .
Residuals 24 0.022667 0.000944
我对F的理解是这样的,某个factor的F就是有多大比例的研究对象的variation可以用
这个参数的variation来解释。从这个ANOVA的结果来看,mix的variation和residual的
variation之差2.45倍。就算忽略这个mix 的影响,整体上的误差也不会很大,所以
aggregate对研究对象的影响更大,而且比mix大得多。
我统计学的不好,说的不对见笑了。 |
|
r*******a 发帖数: 6 | 44 so repeated measures ANOVA is applicable for this analysis ? |
|
w***c 发帖数: 68 | 45 Yes, you did.
Putting ANOVA aside, what test could I use to compare within-group
variability? Is there a variance equivalent to multiple comparison of group
means?
thank you! |
|
h*********r 发帖数: 25 | 46 想用Sigmaplot12 two-way ANOVA分析数据,结果数据输入方法不对,请问应该怎样输
入数据?
谢谢! |
|
d*********k 发帖数: 1239 | 47 A B Y
-1 -1 3.2
1 -1 2.6
-1 1 3.5
1 1 3
如果A, B是factor 然后怎么做ANOVA分析啊?
如果加上了interactoin项,就更不知道咋弄了?怎么办啊?
Analysis of Variance for Y, using Adjusted SS for Tests
Source DF Seq SS Adj SS Adj MS F P
A 1 0.30250 0.30250 0.30250 **
B 1 0.12250 0.12250 0.12250 **
A*B 1 0.00250 0.00250 0.00250 **
Error 0 * * *
Total 3 0.42750
** Denominator of F-test is zero or undefined.
S = *
* NOTE * Could not graph the... 阅读全帖 |
|
m********9 发帖数: 208 | 48 Variance components, a classic book written by Searle, S. R. If you are
interested in the ANOVA analysis for some experiments, the book 'A first
course in design and analysis of experiment', which is a free book. |
|
m*****x 发帖数: 10 | 49 unbalanced模型的repeated measures ANOVA
除了SPSS和SAS,还有哪家可以实现?
R可以吗?有没有公开的算法可以找到? |
|
|