Do the units matter? Validity of Intensity Based Meta-Analysis in the presence of unit mismatches

Abstract : While most neuroimaging meta-analyses are based on peak coordinate data, the best practice method is an Intensity-Based Meta-Analysis that combines the voxel-wise effect estimates and their standard errors (E+SE's, aka COPE & sqrt-VARCOPE) from each study [5]. Such analysis assumes that the E+SE's are expressed in the same units across studies. In practice the observed units of the E+SE's will depend on 1) the method used to scale the BOLD signal; 2) the scaling of the explanatory variables in the design matrix; 3) the scaling of the contrast vector [4]. An analysis of E+SE's therefore requires knowledge of the data, design and contrast scaling. However those meta-data are rarely shared, leaving the researcher with E+SE's with unknown scaling. Here, we investigate the validity of 5 random-effects meta-analytic approaches in the presence of imperfectly scaled contrast estimates.
Document type :
Conference papers
Complete list of metadatas

https://www.hal.inserm.fr/inserm-01149475
Contributor : Camille Maumet <>
Submitted on : Monday, August 24, 2015 - 3:44:07 PM
Last modification on : Sunday, December 6, 2015 - 1:26:47 PM
Long-term archiving on : Wednesday, November 25, 2015 - 4:29:37 PM

File

Maumet-units-OHBM2015.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inserm-01149475, version 1

Collections

Citation

Camille Maumet, Thomas Nichols. Do the units matter? Validity of Intensity Based Meta-Analysis in the presence of unit mismatches. 21st Annual Meeting of the Organization for Human Brain Mapping (OHBM), Jun 2015, Honolulu, United States. ⟨inserm-01149475⟩

Share

Metrics

Record views

137

Files downloads

89