sdm-help-list
82 Subscribers 
sdm-help-list > question about the principle of SDM
Showing 1-3 of 3 posts
Aug 17, 2014 11:08 AM | Nobody
question about the principle of SDM
Hi,
We are conducting a meta-analyses to compare the brain activity between MDD patients and healthy controls. But we have some problem in understanding the principle of SDM. Here are our two questions:
1、Why did SDM use a random-effects model for the main analysis and a sample-size-weighted model for the meta-regression?
2、why SDM values in the meta-regression should range between -1 and 1?
Could you please answer the questions? Thank you in advance!
Sincerely,
Daisy
We are conducting a meta-analyses to compare the brain activity between MDD patients and healthy controls. But we have some problem in understanding the principle of SDM. Here are our two questions:
1、Why did SDM use a random-effects model for the main analysis and a sample-size-weighted model for the meta-regression?
2、why SDM values in the meta-regression should range between -1 and 1?
Could you please answer the questions? Thank you in advance!
Sincerely,
Daisy
Aug 18, 2014 03:08 AM | Nobody
RE: question about the principle of SDM
Hi
In your previous article (Br J Psychiatry. 2009 Nov;195(5):393-402), i found the answer for my 2nd question:the meta-regression SDM value is derived from the proportion of studies that reported grey matter changes near the voxel, so it isexpected that the values of some of the studies are at 0 or near+1 (instead of beingclose to the line).
However, i still feel confused that, in the ES-SDM, what is meta-regression Effect Size value derived from?
In the method of this article, it was writed that "The potential effect of several relevant sociodemographic and clinical variables is examined by means of simple linear regression,weighted by the squared root of the sample size and restricted to only predict possible SDM values (i.e. from-1 to 1) in the observed range of values of the variable."
Are there some difference between meta-regression in SDM and that in ES-SDM? Specifically, is the meta-regression in the ES-SDM also weighted by the squared root of the sample size and restricted to some range?
Sincerely,
Daisy
In your previous article (Br J Psychiatry. 2009 Nov;195(5):393-402), i found the answer for my 2nd question:the meta-regression SDM value is derived from the proportion of studies that reported grey matter changes near the voxel, so it isexpected that the values of some of the studies are at 0 or near+1 (instead of beingclose to the line).
However, i still feel confused that, in the ES-SDM, what is meta-regression Effect Size value derived from?
In the method of this article, it was writed that "The potential effect of several relevant sociodemographic and clinical variables is examined by means of simple linear regression,weighted by the squared root of the sample size and restricted to only predict possible SDM values (i.e. from-1 to 1) in the observed range of values of the variable."
Are there some difference between meta-regression in SDM and that in ES-SDM? Specifically, is the meta-regression in the ES-SDM also weighted by the squared root of the sample size and restricted to some range?
Sincerely,
Daisy
Aug 20, 2014 01:08 PM | Nobody
RE: question about the principle of SDM
Dear Daisy
ES-SDM (Eur Psychiatry 2012; 27:605–611) uses standard meta-analytic random-effects models for both the main analysis and the meta-regressions, with values (effect sizes) thus ranging from minus infinity to infinity.
Note that as you say this was not the case in the original SDM method described in (Br J Psychiatry 2009; 195:393-402), but this method should be considered obsolete now.
Also, I strongly recommend the use of the most updated version, currently AES-SDM (Front Psychiatry 2014; 5:13).
With the best wishes,
Joaquim
ES-SDM (Eur Psychiatry 2012; 27:605–611) uses standard meta-analytic random-effects models for both the main analysis and the meta-regressions, with values (effect sizes) thus ranging from minus infinity to infinity.
Note that as you say this was not the case in the original SDM method described in (Br J Psychiatry 2009; 195:393-402), but this method should be considered obsolete now.
Also, I strongly recommend the use of the most updated version, currently AES-SDM (Front Psychiatry 2014; 5:13).
With the best wishes,
Joaquim