JiscMail Logo
Email discussion lists for the UK Education and Research communities

Help for SPM Archives


SPM Archives

SPM Archives


SPM@JISCMAIL.AC.UK


View:

Message:

[

First

|

Previous

|

Next

|

Last

]

By Topic:

[

First

|

Previous

|

Next

|

Last

]

By Author:

[

First

|

Previous

|

Next

|

Last

]

Font:

Proportional Font

LISTSERV Archives

LISTSERV Archives

SPM Home

SPM Home

SPM  1999

SPM 1999

Options

Subscribe or Unsubscribe

Subscribe or Unsubscribe

Log In

Log In

Get Password

Get Password

Subject:

Re: combined analysis vs separate analysis

From:

Ian Nimmo-Smith <[log in to unmask]>

Reply-To:

Ian Nimmo-Smith <[log in to unmask]>

Date:

Fri, 19 Mar 1999 15:54:57 +0000

Content-Type:

text/plain

Parts/Attachments:

Parts/Attachments

text/plain (101 lines)

Dear Kota and Karl,

> Dear Kota,
> 
> > I am now confused by the different results of the following two types of
> > analyses:
> > 
> > (a) within-subject analysis performed on the pooled data from 10 subjects,
> >     setting the contrast [-1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
> >                                     [0 0 -1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> > 0] ...and so on
> > 
> > 
> > (b) a single-subject analysis performed on the data from the
> > corresponding one subject,
> >    setting the contrast [-1 1]
> > 
> > As far as I have experienced, activation revealed by (a) is smaller
> > than that revealed by (b).  In an extreme case, significant activation
> > observed in analysis (b) completely disappears when I switch to
> > analysis (a) even though the same significance level (and the same
> > subject of course) is chosen.
> > 
> > Here, my questions:  (1) Does this indicate the inhomogeneity of error
> > variance across those 10 subjects?
> 
> It could do.  This is a good point.  If some subjects had very high
> error variance then this would render some subject-specific contrasts
> less sensitive.
> 

But as Kota presents it, it's a case of seeing more activation (higher 
sensitivity?) generally in the separate subject specific analyses than 
in the group analysis, indicating lower error variance for individual 
subjects than in the combined subjects analysis.

As I understand the way SPM goes about its GLM business, Kota's single 
subject analyses use purely within-subject error terms (essentially the 
replications-within-conditions variance). The combined subjects 
analysis uses a mixed within- and between- subjects error term 
(essentially it pools all the above single-subject analysis error terms 
and adds in the subjects-by-conditions interaction variance).

There are therefore two potential types of inhomogeneity: (1) different 
within-subject error terms, which is what Karl is addressing, and (2) 
different (typically larger) subject-by-conditions interaction variance 
than within-subject variance. Type (1) inhomogeneity will be greater 
sensitivity from subjects than from others; type (2) can mean less 
sensitivity from the group analysis than from individual subjects.

> > (2) If so, is it inappropriate to pool those data together and perform
> > multi-subject analysis? Or can I still believe that GLM is robust
> > against such violation of homogeneity?
> 
> The GLM is robust to violations of homogeneity but it may not be
> sufficiently robust in your case.  I think this reduces to an empirical
> question.  I would compare the subject-specific SPMs using separate and
> combined models.  If you are right some SPMs will show too much
> activation and others too little when comparing the SPMs of each
> subject.  One simple way of checking, anecdotally, for heteroscedasticity
> of this sort is to simply look at the adjusted data (using spm_plot in
> results).  Subjects with high levels of error variance should be
> apparent on visual inspection.  If you have not done so already I would
> use proportional scaling for global normalization.
> 

It is arguably plausible that the GLM is robust against type (1) 
inhomogeneity, but less plausible in relation to type (2). The random 
effect model analysis sets about precisely uncoupling these two sources 
of variance. However, in its classical form it discards the smaller 
replications-within-condition-within-subject variance and leave just 
the subjects-by-conditions variance for assessing the significance of a 
condition contrast either for the whole group or for a single subject. 
This of course means a loss of sensitivity, which is further compounded 
by losing all the degrees of freedom arising from the replications.

An extreme version of Koda's anomaly is that one could have a voxel 
where each individual subject shows a significant task vs. rest 
activation effect, but that none of these effect survive as a 
significant single-subject contrast in the combined subjects analysis - 
and possibly even the group task vs. rest effect is sufficiently 
diluted to lose significance.

> With best wishes - Karl
> 

Best wishes,

Ian
---
Ian Nimmo-Smith
MRC Cognition and Brain Sciences Unit
15 Chaucer Road
Cambridge CB2 2EF





%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Top of Message | Previous Page | Permalink

JiscMail Tools


RSS Feeds and Sharing


Advanced Options


Archives

December 2019
November 2019
October 2019
September 2019
August 2019
July 2019
June 2019
May 2019
April 2019
March 2019
February 2019
January 2019
December 2018
November 2018
October 2018
September 2018
August 2018
July 2018
June 2018
May 2018
April 2018
March 2018
February 2018
January 2018
December 2017
November 2017
October 2017
September 2017
August 2017
July 2017
June 2017
May 2017
April 2017
March 2017
February 2017
January 2017
December 2016
November 2016
October 2016
September 2016
August 2016
July 2016
June 2016
May 2016
April 2016
March 2016
February 2016
January 2016
December 2015
November 2015
October 2015
September 2015
August 2015
July 2015
June 2015
May 2015
April 2015
March 2015
February 2015
January 2015
December 2014
November 2014
October 2014
September 2014
August 2014
July 2014
June 2014
May 2014
April 2014
March 2014
February 2014
January 2014
December 2013
November 2013
October 2013
September 2013
August 2013
July 2013
June 2013
May 2013
April 2013
March 2013
February 2013
January 2013
December 2012
November 2012
October 2012
September 2012
August 2012
July 2012
June 2012
May 2012
April 2012
March 2012
February 2012
January 2012
December 2011
November 2011
October 2011
September 2011
August 2011
July 2011
June 2011
May 2011
April 2011
March 2011
February 2011
January 2011
December 2010
November 2010
October 2010
September 2010
August 2010
July 2010
June 2010
May 2010
April 2010
March 2010
February 2010
January 2010
December 2009
November 2009
October 2009
September 2009
August 2009
July 2009
June 2009
May 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
2006
2005
2004
2003
2002
2001
2000
1999
1998


JiscMail is a Jisc service.

View our service policies at https://www.jiscmail.ac.uk/policyandsecurity/ and Jisc's privacy policy at https://www.jisc.ac.uk/website/privacy-notice

Secured by F-Secure Anti-Virus CataList Email List Search Powered by the LISTSERV Email List Manager