Statistical formula for calculating 95% CIs from two geometric means

[Sal Watts]

Hello,

I have the following data:

Group 1: Mean = 8985, confidence intervals = 5705–14150
67 people

Group 2: Mean = 7622, confidence intervals = 5486–10591
132 people

I do not have any individual data that made these means. Only the above data.

Overall mean: 8056.

Overall mean has been calculated as follows:

. di 67*log(8985)+132*log(7622)
1789.8427

. di 1789.8427/199
8.9941844

. di exp(8.9941844)
8056.0964

.

I need to calculate the overall CIs.

Is it the same method?

So for example the lower overall CI calculated as:

. di 67*log(5705)+132*log(5486)
1716.0036

. di 1716.0036/199
8.6231337

. di exp(8.6231337)
5558.7787

.


Or is this incorrect?

Thanks

[William Lisowski]

This question was earlier posted and discussed at

https://www.statalist.org/forums/for...eometric-means

[Sal Watts]

Yes indeed. I checked with the FAQ and it advised not to bump a thread, but rather improve the question. As you highlighted in my previous thread, this was statistical methodological issue so I have reframed my question in line with that, and have added in some of the workings I have attempted.

If anyone could possibly offer help/advice I would be very grateful.

Kind regards.

[William Lisowski]

From the FAQ you read

If you have what is essentially a new and different question, it is generally best to start a new thread. A cross-reference to the previous thread is helpful if there is a relation.

This question is not so new and different from your previous question that those who read this this would not benefit from seeing what you've been told previously before they set about framing an answer. Hence my post #2, for the benefit of those who might be able and inclined to help here.

[Sal Watts]

Okay thanks for clarifying.

If anyone could possibly offer help/advice I would be very grateful.

Kind regards

[Jeff Wooldridge]

Sal: You’re willing to assume independence across the two samples, correct?

[Jeff Wooldridge]

Sal: You’re willing to assume independence across the two samples, correct?