CI95% around geometric mean output in report table
#1
Posted 19 March 2010 - 03:41 PM
CI Geo 95% lower or upper
CI 95% lower (or upper) Geo mean
My question is 1) which of the following two (lower and upper) is what I need and 2) why are they different?
When I further searched the HELP in phoenix, it gave the following definitions but I am still confused.
CI GEO X% Lower: Lower limitfor the geometric mean based on the chosen confidence interval.
CI GEO X% Upper: Upper limit for the geometric mean based on the chosen confidence interval.
CI X% Lower GEO Mean : Lower limit of an X% confidence interval for the geometric mean.
CI X% Upper GEO: Mean Upper limit of an X% confidence interval for the geometric mean.
Need to point out that there is a typo in the definition for "CI GEO X% Upper", it is not "upper limit for the geometric mean..." instead of "lower limit for..."
#2
Posted 19 March 2010 - 04:01 PM
Dear Zan,
Thank you for pointing out the typo in our help. The section about Confidence Intervals will be revised to include the formulas. I have pasted below the defintiions for the CIs you were inquiring about.
It is hard for me to comment on what you need as I am not sure what you want but the most commong reported CI on Geometric mean is the Lower and Upper CI on the Geometric MEAN. See below for definitions.
CI GEO % Lower: The lower confidence interval that contains X% of the logs of the data, back-transformed to original scale. exp(logmean – t(alpha/2) * logSD)
CI GEO % Upper: The upper confidence interval that contains X% of the logs of the data, back-transformed to original scale. exp(logmean + t(alpha/2) * logSD)
CI% Lower GEO Mean: The lower confidence interval that contains the geometric mean with X% probability. exp(logmean – t(alpha/2) * (logSD/sqrt(n)))
CI% Upper GEO Mean: The upper confidence interval that contains the geometric mean with X% probability. exp(logmean + t(alpha/2) * (logSD/sqrt(n)))
GEO Lower XSD: Lower range determined by subtracting 'x' log standard deviations from the log-mean, back transformed to original scale.
Exp(Geo Mean – XSD of Log) where X = the number of Standard deviation chosen
GEO Upper XSD: Lower range determined by adding 'x' log standard deviations from the log-mean, back transformed to original scale
Exp(Geo Mean + XSD of Log) where X = the number of Standard deviation chosen
Hope this helps,
Ana
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users