Date:         Fri, 27 Jun 2003 10:18:19 +0200
Reply-To:     Rita Clivio <Clivio@databank.it>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Rita Clivio <Clivio@databank.it>
Subject:      Confidence interval of Composite Estimates
Content-Type: text/plain; charset="iso-8859-1"

Dear Listers I'd ask your opinion about a statistical problem.

I make an example to clarify the question.

I wish have an estimate on oil consumption (in litre)  in a country; I know that in some regions people use more oil than in others.
In any case, I interview a national sample, asking oil monthly consumption.
I could:
(a) keep the estimate obtained from the total sample (i.e. average consumption multiplied by consumers percentages ? I can also calculate the confidence interval at a confidence level)
(b) sum the estimate obtained by each region.

Using the second method how I can calculate the confidence interval or the estimate precision of the resulting sum ? Could I think that with (b) I have a compensation between each error estimate ? And which method is better according to you?

Thank you in advance for your kind contribution.

Rita