Monday, December 18, 2006

Another nasty question

I think the single most frequent question during statistical consultancy here at Foulum is:

How do I calculate the standard eror on the original scale when I have made an analysis on transformed values.

Back in 1998 I wrote a short note adressing this problem together with Asger Roer Pedersen (now an employee at the National Envirenmental Institute).

Unfortunately, people have trouble finding it, so just to refresh

Jørgensen, E. & Pedersen A.R. (1998 ). How to obtain those nasty standard errors from transformed data - and why they should not  be used. Biometry Research Unit - Internal report 7. Danish Institute of Agricultural Sciences. pp 20

Or, via my publication page: Erik Jørgensen's Publications

No comments: