Wie lügt forscht man mit Statistik? Auch tote Fische haben Gefühle.
Im “Journal of Serendipitous and Unexpected Results” ist die Arbeit “Neural Correlates of Interspecies Perspective Taking in the Post-Mortem Atlantic Salmon: An Argument For Proper Multiple Comparisons Correction” von Craig Bennett, Abigail Baird, Michael Miller and George Wolford erschienen (siehe auch “Großhirn-Voodoo” auf Spiegel-Online und “Ein Fisch schaut in die Röhre” in der SZ).
Es geht um bildgebende Hirnforschung, speziell um die Hirntätigkeit des portmortalen atlantischen Lachses, und um Statistik.
Nun bin ich weder Statistiker noch kann ich beurteilen, ob in der bildgebenden Hirnforschung wirklich so gearbeitet wird, wie es der Artikel nahelegt. Aber das Experiment ist einfach zu schön, als daß man es hier nicht erwähnen könnte.
Zunächst zur Methodik:
One mature Atlantic Salmon (Salmo salar) participated in the fMRI study. The salmon measured approximately 18 inches long, weighed 3.8 lbs, and was not alive at the time of scanning. It is not known if the salmon was male or female, but given the post-mortem state of the subject this was not thought to be a critical variable.
The task administered to the salmon involved completing an open-ended mentalizing task. The salmon was shown a series of photographs depicting human individuals in social situations with a specified emotional valence, either socially inclusive or socially exclusive.
The salmon was asked to determine which emotion the individual in the photo must have been experiencing.
The photo stimuli were presented in a block design, with each block consisting of four photos presented individually for 2.5 seconds each (10 seconds total) followed by 12 seconds of rest. A total of 12 blocks of photo presentation were completed with 48 photos presented during the run. Photos were presented with the experiment-scripting program Psyscope (Cohen et al.,1993) and advanced by a TTL voltage trigger from the scanner. Total scan time for the task was 5.8 minutes, with 140 acquired image volumes.
Voxelwise statistics on the salmon data were calculated through an ordinary least-squares estimation of the general linear model. Predictors of the hemodynamic response were modeled by a boxcar convolved with a canonical hemodynamic response function. A temporal high pass filter with a cutoff period of 128 seconds was included to account for low frequency drift in the functional imaging data. No autocorrelation correction was
Die hier ausgelassenen Details über verwendete Technik und Programme kann man im Original nachlesen, ebenso wie die quantitativen Ergebnisse der Untersuchung. Jedenfalls wurden in einem Cluster in einem Hohlraum des Lachshirnes mehrere aktive Voxel beobachtet:
Either we have stumbled onto a rather amazing discovery in terms of post-mortem ichthyological cognition, or there is something a bit off with regard to our uncorrected statistical approach.
beginnt der Diskussionsteil und diskutiert dann mögliche Lösungen des “multiple comparisons problem in functional imaging”.
There have been several in-depth articles regarding the multiple testing problem in neuroimaging, but a sizable fraction of published research still report results using uncorrected statistics.
The control of false positives is not a matter of difficulty, as all major analysis packages for fMRI include some form of multiple comparisons correction. Rather it seems to be the case that investigators do not want to jeopardize their results through a reduction in statistical power. While we must guard against the elimination of legitimate results through Type II error, the alternative of continuing forward with uncorrected statistics cannot be an option.