W**********E 发帖数: 242 | 1 Likelihood ratio test example:
H0: var1=0 likelihood ratio chi-square test statistics is 6.6976 with df=1.
So 1-pchisq(6.697,1)= 0.0093.
However, this example gets a p-value of 0.0047 by dividing 0.0093 with 2.
I am not sure why we need to divide a 2 here. Variance is always positive
and Ha: var1>0 justifies a one tailed test only, right?
Any help will be appreciated. |
s*r 发帖数: 2757 | 2 devide 2 to convert the 2 sided test to 1 sided test
but the LR test on the border of the region can be tricky
the tests may be wrong |
W**********E 发帖数: 242 | 3 I am not sure I follow this " convert 2 sided test to 1sided test" here.
Pr(Y>test_statistics) is one sided p-value only, right?
if we want a two-sided p-value, we should have a 2*Pr(Y>test_statistics)
under a symmetric distribution.
So I am very confused with the example. Either they report 0.0093 as p-value
for one sided test Ha: var1>0. or 2*0.0093=0.018 as a two sided p-value for
Ha: var1 ne 0. Why divide 0.0093 by 2 here? Plus I do not think it is necessary to do a two tailed test for varia
【在 s*r 的大作中提到】 : devide 2 to convert the 2 sided test to 1 sided test : but the LR test on the border of the region can be tricky : the tests may be wrong
|
x*********i 发帖数: 55 | 4 Maybe you're testing the significance of one random effect in a linear mixed
model, the distribution of the lrt under the null hypothesis var1=0 is a 50
p-value by 0.5( approximate value). See reference: Self and Liang (1987). |
f*******r 发帖数: 257 | 5 I agree with you on that, no need to divide by 2.
value
for
necessary to do a two tailed test for variance=0
【在 W**********E 的大作中提到】 : I am not sure I follow this " convert 2 sided test to 1sided test" here. : Pr(Y>test_statistics) is one sided p-value only, right? : if we want a two-sided p-value, we should have a 2*Pr(Y>test_statistics) : under a symmetric distribution. : So I am very confused with the example. Either they report 0.0093 as p-value : for one sided test Ha: var1>0. or 2*0.0093=0.018 as a two sided p-value for : Ha: var1 ne 0. Why divide 0.0093 by 2 here? Plus I do not think it is necessary to do a two tailed test for varia
|
W**********E 发帖数: 242 | 6 You are right with testing random effects in a linear mixed model.
Actually it is an example on Page60 from SAS system for mixed model-1st
edition. I read the self and liang's paper and emailed author of the book. He gave the same answer.
Thanks for the help. It is a good thing to know
mixed
50
corresponding
【在 x*********i 的大作中提到】 : Maybe you're testing the significance of one random effect in a linear mixed : model, the distribution of the lrt under the null hypothesis var1=0 is a 50 : p-value by 0.5( approximate value). See reference: Self and Liang (1987).
|