Here’s another installment of PubPeer Selections:
- A reminder that correlation is not causation, using a paper from Science as the basis for “a meta-rant.“
- Gel splicing used to be OK, right? So all those older papers were just following standard operating procedure. Uh, no, says a commenter.
- Commenters question figures in a Cancer paper, and authors respond by posting raw data. “Thanks. That makes it clear.”
- One of the researchers who’ve been questioning the “stripy nanoparticles” literature post the reviews of their PLOS ONE paper on the subject, — and discuss their unsuccessful efforts to get Wiley to allow them to publish a figure they were critiquing.
- A commenter notes that the authors of a Human Nature paper on whether black women in the U.S. “experience stress-related accelerated biological aging” found conclusions opposite to those of other studies.
We keep hearing that correlation does not equal causation. How, exactly, does one define causation ?
For sure, I do not have “the definition” of causation. To me, reasoning or physical laws often provide hints of probable (or improbable) causal relationships. Mathematically speaking, a change in the dependent variable can be caused by a change in the independent variable. Then we have a causal relationship between these two of whatever kind. There is one famous example: if one correlates the population of storks in Europe with the birth rate from 1960 to 1990 or so, one finds a near-perfect correlation. Of course, there is no causation.
Statistically, the most widely used Pearson correlation only tests for linear dependence between variables. If there exists a non-linear relation, that correlation is useless.
There is most likely a much better and more thorough definition, which I would like to hear, too! 🙂
Causation is not trivial to establish, and thus the rant on PubPeer about a group that casually ascribes causality. Epidemiologists have been working hard at it for decades, and have guidelines (such as Koch’s postulates) which are well reviewed by the late Sir Richard Doll, see for example
Proof of Causality: Deduction from Epidemiological Observation
Doll, Richard.
Perspectives in Biology and Medicine, Volume 45, Number 4,
Autumn 2002, pp. 499-515 (Article)
Published by The Johns Hopkins University Press
DOI: 10.1353/pbm.2002.0067
available from e.g.
http://www.stats.ox.ac.uk/~snijders/Doll2002.pdf
Larson et al. do have more work to do to support their claim, as the cogent ranter points out.
Regarding gel splicing, I have a query about a paper that was published in Dec 2013 in Korean Journal, J Plant Biotechnol: http://www.koreascience.or.kr/search/articlepdf_ocean.jsp?url=http://ocean.kisti.re.kr/downfile/volume/kspbt/SMSMCM/2013/v40n4/SMSMCM_2013_v40n4_192.pdf
1) Is the banding/gels in Fig. 2 acceptable, especially the Southern blot?
2) All gels in figures 3 and 4 appear to be spliced. Is my assumption correct? If no, please advise.
Also on pubpeer, the Chancellor of the University of Illinois at Urbana-Champaign is accused of triplicate publication with data transcription errors: https://pubpeer.com/publications/2EE674599778410770DFA3BA3A02F8
What’s more serious than the self-plagiarism is that in one of these papers the co-authors have been removed–it is a ‘single author’ paper. What’s worse, making up multiple versions of the same paper, or not giving credit to your co-authors?
Probably the latter. And now, according to a pubpeer comment on another paper by the same authors, it appears that neither behaviour was a once-off:
https://pubpeer.com/publications/72872078379F852062CEC8ADA0939A#fb14353
In fact, the story and the case study are really important. And the person who made the entry at PubPeer did well to maintain a tone-neutral account of the facts, which will win hearts and minds over time. However, I am concerned about the comment at the end: “Is this how one gets to be Chancellor of a major US research university?” This is a value judgment. If PubPeer is to lead, then it must lead by example, of continue to face increasing legal threats. May I suggest, instead, that PubPeer be used to add strictly bland and cold-hard facts but that RW and other relevant blogs, which are meant to stimulate discussion and personal opinions, be used to post questions regarding the link between the errors and the position, salary, benefits, etc of the person involved. By the way, now that there are two entries at PubPeer on the Chancellor of the University of Illinois at Urbana-Champaign, I think an anonymous e-mail to a few dozen University of Illinois at Urbana-Champaign heads of department, questioning his position, and supported by links to the relevant RW and PubPeer pages as evidence, would perhaps start the “desired” ball rolling.
And more:
https://pubpeer.com/publications/0DE4F14CB64C7EC4E36612090CC83F#fb14449
That “Meta” critique is nonsense. The guys spends a lot of time saying the data is great, slightly mis-interprets a result, and then disagrees, perhaps reasonably, with the conclusion. Then, the reviewer goes off the rails to suggest RETRACTION, for a disagreement about INTERPRETATION? Back in the day, you actually did new experiments when you thought someone was wrong. Nowadays, you just post some on-line rantings and consider your intellectual contribution to be somehow meaningful.