That paper with the ‘T’ error bars was just retracted | Hacker News Reader

That paper with the ‘T’ error bars was just retracted

retractionwatch.com

58 pointsadunk29 comments

> Moreover, it has been noted that the authors state that “no datasets were generated or analyzed during the current study” which is contradictory to the study described.

Is peer review even worth anything any more? Is journal publication actually the scam my academic friends have been moaning about for years?

it does, but we would have to spend a lot more to get better papers.

and papers are not the important thing in science. (it's unfortunate, but not really surprising.)

nowadays science is more like an idea generator for startups and other researchers/scientists.

Hindawi journals are borderline predatory and many have no real peer review. See also MDPI - don't trust a thing you see in either unless someone outside of that system has vetted it.

The question the editors should be asking of their peer reviewers is: why was neither this, nor any of the other glaring errors spotted during the process of peer review, a process literally designed to catch things such as this?

This is why I think that peer reviewers should be deanonymized post-review, to allow for accountability and open conflict-of-interest investigations.

It's obvious the bar plot didn't have error bars when it was submitted to the journal. One of the reviewers must have complained about the lack of error bars. So the authors thought "wtf are error bars?", googled, and added some T's...

Deanonymized peer review has no chance of working since identified reviewers would fear retaliation for rejecting someone's manuscript.

>Deanonymized peer review has no chance of working since identified reviewers would fear retaliation for rejecting someone's manuscript.

Rejecting would not be deanonymized, only accepting something with glaring mistakes like here.

The real crime in that chart is that the 'blank' bar is salmon, and the 'pink' bar is yellow.

The interesting part is that, technically, a correctly placed T could actually represent a correct error bar. I mean, they are obviously not (constant spacing), but the fact that a T is used is, by itself, curious, but not wrong.

> a correctly placed T could actually represent a correct error bar

In the way correctly-placed periods could represent a dot plot.

ASCII plots.

There's a few places where its still worthwhile. Low CPU load ASCII spectrum analyzers are a great thing in some applications.

I think the whole chart was made in a drawing program like MS paint or Visio. The spacing between the columns isn’t quite even as would be the case with any software charting library, especially noticeable for the gaps left and right of the blue bar.

Also this quote is amazing

    “the authors state that “no datasets were generated or analyzed during the current study””

Does peer review actually catch a large number of these mistakes, making this an unfortunate rare incident?

Yes and no. It depends on the journal, editor, manuscript, authors.

Post publication peer review will always be much more comprehensive due to the number of eyes reading.

lol if you zoom in , it's obvious. not even trying that hard

Yeah sure junk peer reviews, publisher should have noticed, etc.

Is it not equally, or more, terrifying that an academic sat down and deliberately did this?

I don't get it. They put a T on top? Oh it's supposed to be a line.

Error bars are not a fixed size while the capital T was added without regard for the actual “margin”. Thus making the error bars have 0 value (as they are made up and added for visual appeal it seems)

I thought it was just a label at first. That's funny.

(December 22, 2022)

This is somewhat offtopic, but the Retraction Watch twitter account shared this so many times since December 22, that I had to unfollow them. Now that I see the same article in HN, I’m convinced that the universe is trolling me.

Oh wow didn’t realize from the headline it was a literal uppercase Tee character being used that’s pretty funny (from my perspective)

I wonder if/how quickly it would have been noticed with a 'better' font?

The serifs give it away don't they? The space as it floats over the bar a close second.

If it was some font that looked like a normal error bar, a pretty crap font for proper use probably, just two perpendicular lines, the only clue being that they were all the same size...

Showing up in a text search for "T" would probably also be a bad sign.

The serifs certainly are a dead giveaway, however, there is different amounts of spacing between the top of the bar and the bottom of the character. Additionally, the gray one is visibility off center even in the small image - I'm sure the other ones too.

If they just did an 'add error bars' to Excel and had them be any value, it wouldn't have been discovered (well, that wouldn't have been the issue - the other issues with lack of data and such still would have been present).

Seriously, I can't stop laughing at this.

At least most junk science that gets submitted is done by people who are smart enough to make a chart in Excel.

But when authors don't even know how to put actual falsified error bars (of different sizes) into a chart and just draw identical letter T's on top instead... the incompetence doesn't even seem that criminal anymore, just comical.

Like a bank robber using a banana instead of a gun.

(The journal not catching this before publishing it, however... yikes. Nothing funny there.)

I'm convinced this guy was just trolling, similar to the Sokal hoax: https://en.wikipedia.org/wiki/Sokal_affair

Reply on news.ycombinator.com