On Mon, 23 Nov 2009 11:11:37 +0100, Christoph Hofstetter
<[log in to unmask]> wrote:
>Dear SPM list
>
>I am confused a bit about choosing t-contrast weights on the first
>level. Assuming having two conditions: contrasts [1 -1] and [10 -10]
>produce the same t-value on the first level. However, when performing
>a one sample t-test on a group of each contrast, the significance
>differs (higher multiplier -> higher significance).
That shouldn't happen.
>Maybe related to this:
>When I load the con_NNNN.img images using spm_vol and spm_read_vols
>the corresponding images seem to be equal, however other software
>provides different gray level values for them.
No, that _should_ happen. The higher-multiplier con* images will be bigger by
the multiplier. But second-level statistics should be unaffected.
>So I am wondering how to set the weights in order to have valid second
>level statistics? Which might be especially important when contrasts
>must be balanced e.g. [2 -1 -1] vs [1 -0.5 -0.5] vs [0.5 -0.25 -0.25]
>etc.
>
>The versions I am using:
>Matlab 7.7.0.471 (R2008b)
>SPM8 3408
>
>Thanks in advance to any help
>Cheers
>Christoph
>
>
>--
>Christoph Hofstetter, MSc in Biology
>Laboratory for Neurology and Imaging of Cognition
>Dept of Neurosciences
>University Medical Center (CMU)
>1 Michel-Servet - 1211 GENEVA - CH
>http://labnic.unige.ch
|