With respect, the natural log transformation is the easiest of all to
interpret. It is made even easier by multiplying by 100, i.e.
transforming X to 100 ln X, in which case the corresponding
regression coefficient is the change in outcome per 1% change in X.
My paper (Cole TJ. Sympercents: symmetric percentage differences on
the 100 loge scale simplify the presentation of log transformed data.
Statistics in Medicine 2000;19:3109-25) explains why this is. Email
me if you'd like a PDF of it.
Tim Cole
At 22:19 +0100 27/2/06, BXC (Bendix Carstensen) wrote:
>If you take the log10 of an explanatory variable, the estimated effect
>will be the effect of a 10-fold increase in the original variable, if
>you take the log2, it will be the effect of a doubling, and if you take
>natural log it will the effect of a 2.718282-fold increase in the
>original vaiable. The latter is difficult to sell in publications...
>
>Best,
>Bendix
>----------------------
>Bendix Carstensen
>Senior Statistician
>Steno Diabetes Center
>Niels Steensens Vej 2
>DK-2820 Gentofte
>Denmark
>tel: +45 44 43 87 38
>mob: +45 30 75 87 38
>fax: +45 44 43 07 06
>[log in to unmask]
>www.biostat.ku.dk/~bxc
>----------------------
>
>
>
>> -----Original Message-----
>> From: A UK-based worldwide e-mail broadcast system mailing
>> list [mailto:[log in to unmask]] On Behalf Of Raphael Fraser
>> Sent: Monday, February 27, 2006 5:08 PM
>> To: [log in to unmask]
>> Subject: Cox Regression/ Survival Analysis
>>
>>
>> I am performing a cox regression but have logged one of my
>> independent variable. How can I interpret this variable?
>>
> > Raphael
--
[log in to unmask] Phone +44(0)20 7905 2666 Fax +44(0)20 7905 2381
Paed Epidemiology & Biostats, Institute of Child Health, London WC1N 1EH, UK
|