I would have to dissagree with Pierre's difinitive "You can't". If we make

a probability statement such as Prob(ll < mu < ul) = 90%, (ie construct a

confidence interval for the block mean of log transformed data) , then we

are able to also make the probability statement Prob(f(ll) < f(mu) < f(ul))

= 90% provided that the function f is a one to one function. In the case of

log transformed data, the function we want here is the exponential and as

Pierre notes we may corect for bias using s^2/2 or this can be refined

slightly as mentioned by Cressie 1990 where the bias correction is

porportional to s^2/2. If we do not correct for bias (ie use just the

exponential function, then we have a confidence interval for the median

exp(mu) of the lognormal distribution.

In general the idea of back transforming the confdence limits is a fairly

standard technique throughout the statistical literature.

+---------------------------------------+

|Western EcoSystems Technology, Inc. |

|2003 Central Ave. |

|Cheyenne, WY 82001 |

|phone: 307-634-1756 |

|fax: 307-637-6981 |

|web:

http://www.west-inc.com/ |

+---------------------------------------+

--

*To post a message to the list, send it to

ai-geostats@....

*As a general service to list users, please remember to post a summary

of any useful responses to your questions.

*To unsubscribe, send email to

majordomo@... with no subject and

"unsubscribe ai-geostats" in the message body.

DO NOT SEND Subscribe/Unsubscribe requests to the list!