bacle
Posts:
838
From:
nyc
Registered:
6/6/10


Re: More dumb question on hypothesistesting: Symmetry of Ha.?. Sample size.?
Posted:
Dec 31, 2010 7:55 PM


> Bacle said: Posted: Dec 8, 2010 10:50 AM > > > As far as I know, the alternative Ha dictates the > choice of confidence interval: onesided if Ha is of > the form: Ha: Mu<m or: Ha: Mu>m > And the interval is onesided if the Ha is of the > e form: Ha: Mu =/ m. But I would wait for more > qualified people to reply here if I were you. > ________________ > > My response: > > > If I were you I was very careful in order no to say > nonsense.
But apparently you would not read carefullyenough:
> > a) Where you got the *singular* idea that Ha > determines the *Acceptance* Confidence Interval end > points since the Test Statistics Distribution for H0 > true is the only statistical tool to define the tails > so, the decision points?
I never said, nor implied that. All I said was that the form of Ha (which itself depends on the form of Ho ) indicates whether we will use a 1sided interval or a twosided interval.
> The Test Statistics Distribution for Ha (which never > plays a role in Hypothesis test) leads to evaluate > Beta: the probability of accepting H0 WHEN Ha IS > TRUE, the Type II error. > b) Who did said you that Ha: Mu =/ m which leads > to H0: Mu = m which?s a TWO TAILS *ACCEPTANCE* > INTERVAL not one tail as you wrongly thinks?
Yes, this is a mistake I made. I think if you read my post, it is clear that I meant to say that a statement Ha: Mu=/ m goes together with a twotailed test, given that I said you use a 1sided for Ha: Mu>m or for Ha: Mu<m ; I would not likely have purposefully stated separately that you also use a 1sided for Mu=/m ; if I had truly meant to say that you use a 1sided for Ha:Mu=/m , I would not have stated the three cases separately.
> > Luís A. Afonso

