using differing methods for confidence intervals with differing results

#1
Hello,

Does anybody know the protocol for when you use two different methods to create confidence intervals and get differing results in the test.

I ran a test for a parameter estimation MLE and got different results for my confidence interval. And would like some advice.

For example,

you have a null hypothesis

H0 : m = 20
HA: m is not equal to 20.

Then you use one method for hypothesis testing and you get a confidence interval

(10,15) (reject the null)


Then you use a different method to test a hypothesis and you get;

(15,25) (fail to reject)

What would you do under these circumstances?

e.g Not use multiple tests, or fail to reject because one was failed, yeah I am not sure, hopefully you understand the problem and can help me thanks.
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
Hello,

Does anybody know the protocol for when you use two different methods to create confidence intervals and get differing results in the test.

I ran a test for a parameter estimation MLE and got different results for my confidence interval. And would like some advice.

For example,

you have a null hypothesis

H0 : m = 20
HA: m is not equal to 20.

Then you use one method for hypothesis testing and you get a confidence interval

(10,15) (reject the null)


Then you use a different method to test a hypothesis and you get;

(15,25) (fail to reject)

What would you do under these circumstances?

e.g Not use multiple tests, or fail to reject because one was failed, yeah I am not sure, hopefully you understand the problem and can help me thanks.
Well, ideally you had a protocol draft a priori that you strictly adhere to. What was the second method, OLS, bootstrap, etc.? Also, what is the sample size?
 
Last edited: