Now for some Dialectic. Or how to make a physicist lose it: I think John Ringo (who put this on Facebook) has tried this argument on some of the physicists he writes books with. Carefully. His co author holds a black belt and builds rockets.

NB. I don’t link to the Borg.

Schrödinger’s cat is a thought experiment, sometimes described as a paradox, devised by Austrian physicist Erwin Schrödinger in 1935.[1] It illustrates what he saw as the problem of the Copenhagen interpretation of quantum mechanics applied to everyday objects. The scenario presents a cat that may be simultaneously both alive and dead,[2][3][4][5][6][7][8] a state known as a quantum superposition, as a result of being linked to a random subatomic event that may or may not occur. The thought experiment is also often featured in theoretical discussions of the interpretations of quantum mechanics. Schrödinger coined the term Verschränkung (entanglement) in the course of developing the thought experiment.

The concept was postulated in 1935. It is 2016. That is 81 (human) years after the idea. The average lifespan of a cat is 15 years.

Quantum physics is now solved. The answer is: Dead.

Via Vox, an Ars Technica article on functional MRI: it appears the software is broken. In all three packages commonly used. An article I better share with my neuroimaging colleagues. What I can say is that to date neuroimaging has been intriguing… but of limited clinical value. We can’t make diagnoses on the imaging: it appears that neural networks are more likely to help, but that software is also in evolution and we don’t have something that is reliable and acceptable, yet simple and cheap.

In a socialised health system (which NZ has) it has to be simple and cheap.

The authors started with a large collection of what were essentially controls, randomly chose some to be controls again, and then randomly chose others to be an “experimental” population. They repeated this thousands of times, feeding the data into one of three software packages. The process was repeated with slightly different parameters in order to see how this affected the outcome.

The results were not good news for fMRI users. “In brief,” the authors conclude, “we find that all three packages have conservative voxelwise inference and invalid clusterwise inference.” In other words, while they’re likely to be cautions when determining whether a given voxel is showing activity, the cluster identification algorithms frequently assign activity to a region when none is likely to be present. How frequently? Up to 70 percent of the time, depending on the algorithm and parameters used.

For good measure, a bug that has been sitting in the code for 15 years showed up during this testing. The fix for the bug reduced false positives by more than 10 percent. While good that it’s fixed, it’s a shame that all those studies have been published using the faulty version.

the paper notes that these flaws cannot be removed from the datasets, because the datasets are not standard, and often unavailable. Like all forms of tomograhy, the dataset from any MRI is a large set of numbers that estimate the radiofrequency[1] following shifts in hydrogen atoms in water collected in sensors around the brain which then allow the reconstruction of a matrix of densities. If you have the daw data, you can then reanalyse with better software. But this is not available.

From the authors.

It is not feasible to redo 40,000 fMRI studies, and lamentable archiving and data-sharing practices mean most could not be reanalyzed either. Considering that it is now possible to evaluate common statistical methods using real fMRI data, the fMRI community should, in our opinion, focus on validation of existing methods. The main drawback of a permutation test is the increase in computational complexity, as the group analysis needs to be repeated 1,000–10,000 times. However, this increased processing time is not a problem in practice, as for typical sample sizes a desktop computer can run a permutation test for neuroimaging data in less than a minute . Although we note that metaanalysis can play an important role in teasing apart false-positive findings from consistent results, that does not mitigate the need for accurate inferential tools that give valid results for each and every study.

Finally, we point out the key role that data sharing played in this work and its impact in the future. Although our massive empirical study depended on shared data, it is disappointing that almost none of the published studies have shared their data, neither the original data nor even the 3D statistical maps. As no analysis method is perfect, and new problems and limitations will be certainly found in the future, we commend all authors to at least share their statistical results [e.g., via NeuroVault.org (44)] and ideally the full data [e.g., via OpenfMRI.org ]. Such shared data provide enormous opportunities for methodologists, but also the ability to revisit results when methods improve years later.

Meta analysis is valuable, but only if the base data is valid. Sorting that out, from published papers, is a considerable challenge, one that automated tools such as covidence have not solved. But if you cannot get at the raw data, you cannot correct. The ethics around this are difficult: there are cultures who are reluctant to have their genome and variabtions patented and exploited, and the ever present issue of privacy and confidentiality.

But we are left with irreproducable results, and the logic of a long dead cat.

_________________
1. I memorized the physics of this for my fellowship. That was a decade or so ago, and it was painful then and worse now. Most doctors go “Oooh sparkly pictures”. But they are measuring blood flow. That may have nothing to do with what is being remodeled in the brain as you are reading and typing.