 |
|
| Science Forum Index » Space - Consult Forum » relative risk per standard deviation... |
|
Page 1 of 1 |
|
| Author |
Message |
| hans... |
 Posted: Tue Oct 13, 2009 10:25 am |
|
|
|
Guest
|
Hello,
in terms of predictors for osteoporosis-related fracture I've got the
question to calculate the relative risk per standard-deviation (RR per
SD) of a risk predictor.
I'm not familar with RR per SD.
I would do the following:
In case control study, RR can be approximated by odds ratio.
Odds ratio referring to a change of 1 unit of predictor is calculated
as exponential of estimate of logistic regression (model
disease=predictor). So, when we use standardized predictor (mean=0,
SD=1), RR per SD can be directly estimated by exp(estimate) resulting
from logistic regression.
Is that correct? What can I do if RR-approximation is not valid?
Thanks - Hans |
|
|
| Back to top |
|
|
|
| Ray Koopman... |
| Posted: Tue Oct 13, 2009 10:36 am |
|
|
|
Guest
|
On Oct 13, 1:25 pm, hans <tho... at (no spam) gmx.de> wrote:
[quote:b2f7e8a5ce]Hello,
in terms of predictors for osteoporosis-related fracture I've got the
question to calculate the relative risk per standard-deviation (RR per
SD) of a risk predictor.
I'm not familar with RR per SD.
I would do the following:
In case control study, RR can be approximated by odds ratio.
Odds ratio referring to a change of 1 unit of predictor is calculated
as exponential of estimate of logistic regression (model
disease=predictor). So, when we use standardized predictor (mean=0,
SD=1), RR per SD can be directly estimated by exp(estimate) resulting
from logistic regression.
Is that correct? What can I do if RR-approximation is not valid?
Thanks - Hans
[/quote:b2f7e8a5ce]
Your interpretation of RR/SD sounds reasonable to me,
but try asking at http://groups.google.ca/group/MedStats/ |
|
|
| Back to top |
|
|
|
| Bruce Weaver... |
| Posted: Tue Oct 13, 2009 12:21 pm |
|
|
|
Guest
|
On Oct 13, 4:25 pm, hans <tho... at (no spam) gmx.de> wrote:
[quote:f8f9058eb8]Hello,
in terms of predictors for osteoporosis-related fracture I've got the
question to calculate the relative risk per standard-deviation (RR per
SD) of a risk predictor.
I'm not familar with RR per SD.
I would do the following:
In case control study, RR can be approximated by odds ratio.
Odds ratio referring to a change of 1 unit of predictor is calculated
as exponential of estimate of logistic regression (model
disease=predictor). So, when we use standardized predictor (mean=0,
SD=1), RR per SD can be directly estimated by exp(estimate) resulting
from logistic regression.
Is that correct? What can I do if RR-approximation is not valid?
Thanks - Hans
[/quote:f8f9058eb8]
I agree with Ray that your approach sounds reasonable.
Re your last question, one possibility would be to estimate the RR
from the OR using a method such as that described in JAMA.
1998;280:1690-1691. I have a feeling there could be more recent
papers on this subject though, so you'll want to check on that before
adopting the method described in that paper.
--
Bruce Weaver
bweaver at (no spam) lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM." |
|
|
| Back to top |
|
|
|
|
|
All times are GMT - 5 Hours
The time now is Fri Nov 27, 2009 11:55 pm
|
|